The Impact of SENAI`s Vocational Training

Transcription

IRIBA Working Paper: 05
__________________________________________________
The Impact of SENAI's Vocational Training Programme on
Employment, Wages, and Mobility in Brazil: What Lessons
for Sub Saharan Africa?
Carlos Villalobos Barría and Stephan Klasen1
July 2014
________________________________________________________________
1
Ibero-America Institute for Economic Research, University of Goettingen
ISBN 978-1-910502-04-4
International Research Initiative on Brazil and Africa (IRIBA)
School of Environment, Education and Development, The University of Manchester,
Oxford Rd, Manchester, M13 9PL
[email protected]
www.brazil4africa.org
Abstract
The paper investigates to what extent the Brazilian SENAI system of vocational training could
be a role model for easing the substantial challenges African countries face to tackle rising
urbanization, high youth unemployment, and a skills gap. We first discuss relevant features
of the SENAI and associated training systems as they developed over time. Subsequently,
we show that the SENAI system offers opportunities for further training across the
educational and race distribution as well as how the system does not appear to reach the
poorest parts of the population and leaves women under represented. We then study the
returns of the SENAI and other training systems on labour market outcomes and find that
the S-System promotes employment prospects for all groups and the wage premia are
substantial for young males, but much lower for older workers, and very low for women.
SENAI also appears to promote regional migration.
Keywords: Labour markets, vocational training, school to work transition, Brazil, Africa.
Acknowledgement
This document is an output from a project funded by the UK Aid from the Department for
International Development (DFID) for the benefit of developing countries. However, the
views expressed and information contained in it are not necessarily those of or endorsed by
DFID, which can accept no responsibility for such views or information or for any reliance
placed on them.
2
Summary
High rates of youth unemployment, low levels of general and vocational skills, high rates of
rural-urban migration, and poorly functioning labour market institutions impose great
difficulties for young people to manage the school-to-work transition in Sub-Saharan Africa
In Brazil, the National Service for Industrial Training (SENAI) has existed for decades as the
main building block of the Brazilian S-system of vocational training. It is a private non-profit
organization, managed and led by industrial associations that have historically been
considered the leading institution providing professional skills contributing to improve the
school-to-work transition in Brazil. Here we investigate the lessons of this training service
for Sub-Saharan Africa.
Initially, we focus on the ideological controversy that, over decades, has contextualized the
political debate around vocational training, its degree of complementarity to formal
education, the nature of the participant institutions and their models of financing. Then, we
exploit the PNAD 2007 household survey to evaluate the SENAI in dimensions of
performance, which are relevant independently of the country-specific context. Firstly, we
rely on a probability model to investigate the selectivity process into training. Secondly,
based on inverse probability weighting (IPW) and selectivity-adjusted estimators, we assess
the impact of past training episodes on a variety of labour market outcomes. Additionally,
we use quantile regressions to measure the extent in which the heterogeneity of returns to
vocational training affects the urban-rural and gender earning-gaps. Finally, we use a
switching regression model with endogenous switching in order to investigate the extent to
which the SENAI’s vocational training encourages labour mobility.
We find that SENAI graduates exhibit higher employment and productivity (hourly labour
earnings) levels than non-trained workers. However, the impacts are heterogeneously
distributed and favour primarily young heads of household. Moreover, the SENAI seems to
supply its services without consideration of the race (or ethnic background) of the potential
trainee. The same is found when analysing the geographic distribution of training. It means
that the SENAI also offers opportunities for those living in less populated or less advanced
federal states and that differences in participation almost entirely reflect the observed
differences in population by federal states. This neutrality may play a fundamental role in
Africa given the existence of multiple and competing ethnic/racial groups and it is highly
relevant as Africa continues to urbanize. We argue that the federal organization of the SENAI
contributes to this territorially and racially balanced output. Regarding mobility, even when
there is some evidence suggesting that vocational training may discourage further human
capital accumulation and therefore introducing rigidities into the labour market, we provide
novel evidence that SENAI working graduates are more likely to migrate within the country
relative to other workers and thus, this new evidence counters the claims of immobility.
Besides the promising performance of the SENAI in all above mentioned aspects, the fact
that female SENAI graduates perform worst (relative to men) in most analysed labour
market outcomes, raises the question of what explains the observed gender-related
underperformance and whether this disparity is due to discrimination.
3
Studying the Brazilian experience provides a useful reference when analysing the
particularities of the different vocational training models in Africa as well as a clear
understanding of the factors that have contributed to the success and sustainability of the
SENAI since its foundation in the early 1940s. The potential and limitations of the SENAI (and
S-system as well) originate from its market orientation, its decentralised and self-governing
federal organization, together with its financing structure (to a great extent public money).
Since the SENAI can be just as good as the local employers associations are, the first and
more important lesson is that the success and sustainability of SENAI-type of institutions in
African countries will rely, to a great extent, on the quality of the employers’ associations, a
balanced financing structure providing the right incentives as well as a decentralized
corporative government that ensures the required neutrality in the supply of training.
4
1. Introduction
Job creation in the non-agricultural sector has become crucial for policy-makers as Africa
continues to urbanize. This transition is occurring in the context of a strong demographic
expansion (youth bulge), only modest per-capita economic growth, huge levels of
informality, and a still dominant agricultural sector (Biavaschi et al., 2012).1 In many SubSaharan countries the lack of skills aggravates the situation of many young people without a
post-secondary qualification. Large portions of the rural population thus either remain
trapped in the declining traditional sector, or migrate to urban areas experiencing poor
labour market prospects, resulting in, amongst other undesirable consequences there, such
as poor standards of living, informality, and a rise in criminality (Hove et al., 2013).
In urban areas the scarcity of skills also translates into poor labour market outcomes and a
problematic school-to-work transition (See Garcia and Fares, 2008a); this not only hurts the
affected young people, but also the economy as a whole as the lack of skilled workers is
critical for a country’s productivity, growth, and international competitiveness.
Urban populations in Sub-Saharan Africa already account for about 40 per cent of the
population and are expected to double by 2030 (see Hove et al., 2013 and UNPF, 2007
respectively). The fact that, on average, only about 60 per cent of young individuals have
finished primary education shows that the progressive urbanization of Africa is occurring in
the context of low levels of human capital (Garcia and Fares, 2008b). The urbanization
process is a long-run general trend that will not be reversed. It is driven by structural change,
but was also promoted by the old post-independence policies encouraging large-scale
capital-intensive industries in large cities as well as exchange rate policies that negatively
affected employment in the agriculture sector (Hove et al., 2013; World Bank, 1989).2
Consequently, one of the main challenges for policymakers in the upcoming decades in
Africa will be improving the school-to-work transition. The promotion of diversified skills for
dynamic and growing non-agricultural sectors is likely to play an important role in a
continent where vocational training has only played a negligible role in the development of
skills (DFID, 2007; Oketch, 2007).3
Before turning to the Brazilian experience on vocational training, we briefly provide some
empirical analysis of labour market transitions in selected Sub-Saharan African countries to
highlight the challenges just mentioned.
1
The so-called “youth bulge” (term coined by Assad and Roudi-Fahimi in 2007) is a term used to reflect the fact
that the labour demand is unable to absorb a demography-driven massive supply of unskilled labour.
2
Common features of the decline of the traditional sector are large unemployment numbers,
underemployment, high vulnerability to external shocks and fluctuations in commodity prices (Klasen et al.,
2014).
3
In Africa, the difficulties associated with promoting vocational training are mainly related to the general
preference for general education, high shares of informality and limited institutional support (Oketch 2007;
Atchoarena and Delliuc, 2001; Biavaschi et al., 2012).
5
1.1.
Labour market transitions in Sub-Saharan Africa: The
case of Liberia, Malawi and Zambia
Based on comparative analyses of an ILO-coordinated research programme on the school-to
work transition in Africa including Zambia (Chigunta et al., 2013), Liberia (de Mel et al., 2013)
and Malawi (Mussa, 2013), there are some labour market related commonalities worth
considering.4 For instance, education significantly influences the labour market transition of
the youth with those lacking skills having particular difficulties in finding work. Most youth
rely on informal networks when searching for jobs. Moreover, those working are still in their
labour market transitions, as they have not yet got a suitable job for themselves (stable and
satisfactory). In general, young women have to accept the only work available to them,
which tends to be in low-skilled occupations. Another commonality is that the number of
highly skilled youth workers does not satisfy the labour demand for professional
occupations. In Liberia and Malawi, agriculture is still providing most jobs while in Zambia it
is the service sector. Furthermore, in Liberia and Zambia, countries where youth
unemployment rates are relatively high, the length of time spent as unemployed tends to be
long. Conversely, Malawi has relatively low unemployment rates and a short unemployment
duration but significantly higher rates of underutilization and underemployment.5
Table 1 compares country-level representative statistics on education and indicators of the
labour market performance of youth in Brazil and those from the selected African countries.6
While the education gap between Brazil and Zambia is relatively small, this is more
significant against Liberia and it is immense when comparing Brazil and Malawi. Such
heterogeneity is also present when considering the proportion of the youth participating in
vocational education or training. According to the school-to-work transition survey (STWS) of
2012, in Malawi poor educational achievements go hand-in-hand with an almost nonexistent vocational training system. On the opposite end is Zambia, with more than a fourth
of its young population attending vocational training. In Brazil the proportion reached 8.3%
in 2012.
This comparison suggests that Brazilians exhibit neither the highest employment
(enrolment) to-population ratios nor the lowest labour market exclusion rates
(discouragement, neither in employment nor in education or training NEET). The proportion
of those NEET is in fact more than 5 percentage points higher in Brazil than in Liberia.
However, regarding youth unemployment (relaxed and strict definitions) and
underemployment, Brazil significantly outperforms the selected Sub-Saharan African
countries.7 On average, employed young individuals in Brazil work at least 20% more hours
4
These reports are produced by the Work4Youth project by ILO and The MasterCard Foundation.
Underutilization is defined as youth in irregular employment, unemployment (relaxed definition) plus youth
population neither in the labour force nor in education/training as a percentage of the youth population.
Underemployment is related to the weekly hours of work. In Malawi, 41.3% of the employed youth work less
than 10 hours per week (Mussa, 2013).
6
Figures based on International Labour Organization micro-data. In particular the 2012 school-to-work
transition survey (STWS) and the 2012 labour demand enterprise survey (LDES).
7
Broad unemployment (relaxed definition) is a person that (i) did not work in the reference period, (ii) was
available to take up a job had one been offered in the week prior to the reference period. The strict definition
includes the condition (iii) that the person sought work within the past 30 days.
5
6
per week, and the dispersion of the working time is significantly smaller than in Liberia,
Malawi or Zambia.
Table 1: Labour market related characteristics of the youth population (aged 15-29).
Characteristics (percentage)
Brazil
Liberia**
Malawi
Zambia
Education level
Primary or less
27.29
44.4
84.2
24.2
Secondary level
52.42
45.9
14.4
46.2
Tertiary level
12.03
4.7
1.1
1.8
Vocational training
8.3
5.1
0.2
27.8
Employed
53.93
52.42
66.48
43.49
Only enrolled
23.07
30.15
15.94
28.25
NEET
23.00
17.43
17.58
28.26
Unemployed (strict definition)
11.1
19.3
7.8
17.7
Unemployed (relaxed definition)
19.5
35.0
18.9
38.0
Discouraged
2.62
8.56
5.92
6.21
Hours of work (weekly mean)
38.32
32.27
23.15
29.31
Std. Dev.
16.64
25.34
20.86
24.64
Labour-related shortage*
23.31
14.50
49.69
Source: Author’s calculations based on SWTS (2012) and LDES (2012), ILO. *Here, an enterprise with
labour-related shortage declares that its biggest difficulties are due to low labour quality, and/or lack
of skills, and/or low productivity and/or higher labour costs. LDES (2012) is not representative at the
country level. ** Youth population are those aged 15-35.
7
1.2.
The Brazilian experience: the S-system and the SENAI
Having as a reference the considerable heterogeneity across Sub-Saharan African countries
and the challenges related to their labour market transitions, the major contribution of this
paper is to assess who participates in Brazil’s vocational training S-system (mainly SENAI) as
well as to provide an impact evaluation of S-system training on improving the school-to-work
transition.8 In particular, enquiries about what Africa can principally learn from the National
Service for Industrial Training (SENAI) and from the National Commercial Apprenticeship
Service (SENAC) is of great interest.9 Both institutions are the fundamental building blocks of
the so-called S-System and share the same formal structure (De Moura Castro, 2011). The Ssystem is a collection of nine separate initiatives created progressively over the years aiming
to prepare workers through supplying them with the skills needed in the industry as well as
in other sectors. In this paper, we will focus our analysis on the SENAI’s experience in
supplying skills by supplementing the traditional and persistently underdeveloped schooling
system. This experience can shed light on how to set up a sustainable vocational training
system in Africa as well as answer the question of whether SENAI’s framework holds promise
to address the skills level problem and the difficult school-to-work transition in Africa.10
Based on the Pesquisa Nacional por Amostra de Domicílios (PNAD) household survey, we will
evaluate the performance of the Brazilian vocational education model without having the
possibility to disaggregate the performance of individual institutions of the S-system.11
Nonetheless, it is worth noting that the SENAI and SENAC play a fundamental, dominant, and
foundational role within the S-system. They were both created during the 1940s with their
associated social services SESI and SESC.12
1.3.
Vocational vs. general education
There is an old debate on the proper role of general education and the provision of
vocational education. Vocational education (also known as professional education excluding
8
The S-system consists of the National Service of Industrial Learning (SENAI), the Social Service of Industry
(SESI), the National Service of Business Learning (SENAC), the Social Service of Trade (SESC), the National
Service of Rural Learning (SENAR), the National Service of Learning in Transports (SENAT), the Social Service of
Transports (SEST), the Brazilian Service of Support to Small and Medium-sized Companies (SEBRAE) and the
National Service of Learning of Cooperatives (SESCOOP).
9
Different from the SENAI, in the case of SENAC, employers are loosely organized and students pay for most of
the training, at least in part. Thus it is considered a market-driven organization.
10
In Brazil, low achievements in education are highly persistent when compared to other (at that time)
underdeveloped countries. For instance, statistics on the average years of schooling show that Korea had 3.4
years of education in 1950 and 6.8 in 1973, while Brazil reached 2.1 and 3.8 years respectively (Maddison,
1995). PNAD 2009 data shows that in 2009, almost 60 per cent of the population achieved less than secondary
education.
11
According to Vera (2009), SENAI doubles SENAC in the number of trainees in 2007. However, there are no
consolidated figures regarding the whole S-system. The PNAD is part of a long-standing Brazilian tradition of
policy-relevant data collection (going back to the 1920 Census, still one of most remarkable efforts of its time.)
Almost all African countries could learn from PNAD and its parent organization the IBGE – Instituto Brasileiro de
Geografia e Estatística.
12
The other learning institutions of the S-system are SEBRAE and SENAR, created during the 1980s, as well as
SESCOOP, SENAT, and its social service SEST, which followed in the 1990s.
8
college education) refers to the type of education oriented to, almost exclusively, to
generating skills for the labour market.
The ideological controversy is based on the idea that a vocational training system would
divide the society into two groups: one group of low-skilled workers in activities with low
prestige and another group in occupations associated with high incomes and status. The
reason is that vocational training at the secondary level would reduce incentives to further
human capital accumulation and hence constrain equal opportunity and social mobility
(Fresneda, 2009).
In Europe, OECD (2005) points out that occupational shifts in the industrial sector have
affected over-proportionally those graduates with a vocational education (technical level) as
well as those with an incomplete secondary education, contributing to the segmentation of
the labour force and increasing inequality. In the same line of criticism, Goldthorpe and
Erikson (1992) and more recently Müller and Pollak (2005) claim that low levels of social
mobility in Germany are a consequence of the highly developed vocational training system
(See also Shavit and Müller 1998; Breen and Jonsson, 2007; Müller and Gangl, 2003;
Buchmann and Park, 2009).
On the other hand, arguments favouring the promotion of vocational education are based
on the decoupling that tends to occur between the academic curricula (secondary and
tertiary levels) and the practical skills demanded by the enterprises, which would allow them
to participate in competitive international markets, allowing the economies to transit from
the export of commodities towards a diversified production with high added value (see
Schwartzman and Christophe, 2005). It is also argued that vocational education could
contribute to (re-) incorporating to the labour markets those who dropped out from the
educational system or those with very low formal educational achievements (Forte Aguas,
2011). Schwartzman and Christophe (2005) enumerate problems related to an extensive
emphasis on general education already enounced by Grubb (1985), such as educational
inflation, credentialism, over-education, and the high cost associated with promoting
education amongst low income families in order to maintain its relative socio economic
position. The same author points out that vocational training could improve productivity in
the agriculture sector, thus discouraging rural urban migration, which is relevant as Africa
continues to urbanize.
It is worth noting that the consolidation and enhancement of the vocational education in
Europe was a response to the efforts made by the governments for many years, but
intensified during the 1980s to alleviate the high unemployment rates among the young
(Mizen, 2004). In countries where vocational education (technical and technological levels) is
provided, those who at the most completed secondary education spend, on average, more
time searching for a suitable job and face more difficulties during their school-to-work
transitions than those with vocational qualifications (Müller et al., 2003). As a result, the age
profile of unemployment is less unfavourable for young people in countries with welldeveloped vocational training systems such as Germany and Austria even at times when
overall unemployment rates are similar. Contrary to this, those reaching tertiary education
9
levels are more likely to experience a smoother school-to-work transition, with higher labour
earnings and less unemployment than the previous groups.13
Regarding the structure of the vocational training in Brazil, since 1996 the legal framework
(“Lei de Diretrizes e Bases”) has stated that vocational education courses cannot substitute
secondary education and vocational education can only play a complementary role
(Schwartzman and De Moura Castro, 2013). It consists of professional qualification courses
(of relatively low academic level) that are partially dissociated from the schooling system,
the technical level (secondary equivalent), which can be obtained in an integrated manner,
concomitantly or subsequently to the secondary education, and the technological level
(tertiary education equivalent).14 The latter is a less structured, less bureaucratic, more
flexible in terms of its duration, and diploma-granting schema.
13
See Bratberg and Nilsen, 1998; Mcvicar and Anyadike-Danes, 2000; Lassibille et al, 2001; OECD, CPRN 2005;
Audas et al., 2005. See also Fresneda (2009) for a detailed survey of the literature on the educational
trajectories (vocational vs. general education) in the European continent.
14
See also Schwartzman and De Moura Castro, 2013.
10
1.4.
The origins of the SENAI
During the 1930s, Brazil relied almost entirely on the import of industrial goods. In order to
reduce its dependence on the troubled northern hemisphere, after World War II the country
launched a big effort to industrialize the economy (known as the import-substituting
industrialisation strategy). The post-World War II period also brought ideological
confrontations to all levels of society. Employers assumed an anti-labour position that
antagonised the leftist labour militancy and trade unions from 1945 onwards (Colistete,
2010). Starting in the 1940s, a huge flow of rural-urban migrants, badly assimilated into lowskill occupations (“unlimited” supply of low-skilled labour), encouraged high levels of income
inequality in a context characterized by the neglect of the general education (Cardoso,
2008).15 Proof of this is the fact basic education was only universalized during the late 1990s
the substantial concerns about the quality of this universal basic education (Schwartzman
and Christophe, 2005).
Apart from a deficient schooling system, it was also evident that there was also a lack of
programmes to train skilled workers. The scarcity of skilled workers generated the necessity
of implementing a vocational education model. The question of who was going to
administer the model and what the model’s characteristics would be were posed. Who
would administrate the model was particularly controversial since historically, vocational
education and the schooling system were independent of each other.16
Given the poor performance of the government in promoting and improving the schooling
system, it was not surprising that the private sector aspired to operate the nascent
vocational education system.17 The government finally decided to create the SENAI. It was
the first institution of the forthcoming S-system. The SENAI was inaugurated under the
government of Getulio Vargas in January 1942, following the German and Swiss apprenticetraining models (Wilson, 2006). It was organized at the national and state levels as a private,
non-profit organization, financed, managed and led by the industry (Wilson, 1996; Instituto
de Desenvolvimento da Guanabara, 1970).18 Originally, the SENAI was inspired by the
German and Swiss apprenticeship model of providing training as part of dual system of
formal vocational training in training institutes and on-the-job training in firms. However, the
on-the-job component lost momentum over time due to difficulties in finding firms willing to
offer training positions. In fact, during the 1970s, only a third of all needed vacancies were
available (SENAI-DR/RJ, 1979); consequently the SENAI developed as an institution providing
15
See Ribeiro (2007) for a description of the evolution of the Brazilian education system. Over the past four
decades the average years of schooling remained stagnated below four years.
16
According to De Moura Castro (2011), in Brazil, There is and there was a clear division between the Ministry
of Education and the S-System. In his words, in the overlapping areas, there is an acceptable Pax Romana.
17
During the early 1940s, the Law of Industrial Education allowed the Ministry of Education to create and
provide vocational education in the “Liceu Nacional no Rio de Janeiro” with support from Swiss teachers. The
initiative that excluded the agriculture sector was unsuccessful and therefore cancelled (Schwartzman et al.
2000; Schwartzman and Christophe (2005). Moreover, the conflict between educators and confederations has
some history. For instance, in 1971 the law 5692 reformed the education system, introducing a mandatory
professional qualification as a part of the curricula in the secondary education. However, this element was
already removed in 1982.
18
See Rodrigues (1998) on the origins and the linkages of SENAI to the National Confederation of Industry.
11
training as a stand-alone operation not linked to a particular job in a firm, necessitating
graduates to then find employment upon graduation.
Today, SENAI is considered to be among the best and oldest national training system in the
world (Wilson, 1993), having benefited more than 52 million people since 1942 (SENAI-DN
2012). However, although there has been an impressive expansion of the institution, as well
as a flourishing of other institutions offering vocational training, the fact remains that
education in Brazil has progressively moved towards a system prioritizing the general
(academic) education over vocational education which leads to younger cohorts showing
high education levels and the proportion of youths attending vocational education has
steadily decreased over the past four decades.19
1.5.
SENAI and the industry
During its initial years, the SENAI played a marginal but increasingly important role in
supporting the development of the industry in Brazil. Originally the emphasis was put on
apprenticeships, but in the very beginning, the legislation overregulated the apprenticeship
process making it unattractive to many companies causing the initial emphasis on
apprenticeships to shift towards the provision of vocational training (Schwartzman and de
Moura Castro, 2013).20 In this initial stage, real wages grew slower than labour productivity,
as a consequence of that, the gap between industrial wages and profits increased during the
post-war years. Inevitably, the problem of income inequality encouraged antagonistic labour
relations that created an unfavourable environment for fostering manufacturing quality
standards (Colistete, 2010).
The stagnation of the SENAI ended in 1956, a year in which the number of new enrolments
began to grow exponentially. The number of training centres doubled during the period
1964-1968, reaching about 200 units (Instituto de Desenvolvimento da Guanabara, 1970).
The upsurge experienced by the SENAI over the 1960s can also be also associated with a
state-led industrialization effort known as “Post Import Industrialisation” that had long run
consequences (Baer, 2007). As of 1970, the share of industry in GDP had reached about 35
per cent, having remained stable over the period 1970-2000 (Kniivilä, 2007).
Then, in 1973, the oil crisis precipitated the end of the economic miracle (1967-1973)
causing significant transformation within the industrial sector. The apparent stability of the
industrial share of GDP concealed the fact that the industrial specialization subsequently
moved away from labour-intensive activities and towards capital-intensive and natural
resource-based industries, associated with substantial job losses (Cimoli and Katz, 2002).
New employment possibilities shifted towards the service sector to highly remunerated
occupations as well as to low-productivity activities, disequalizing the labour income
distribution (ECLAC, 2004). Meanwhile, SENAI graduates were unable to find jobs and as a
19
For instance, during the early 1970s, one third of all students were enrolled in vocational education
(Hasenbalg, 2003). On the contrary, less than ten per cent were enrolled in this type of education in 2007
(PNAD, 2007).
20
Consequently, between 1946 and 1960, the number of workers who completed some type of apprenticeship
or similar programme provided by the SENAI never reached two percent of the total industrial workforce in the
state of São Paulo, by far the main industrial centre in Brazil (Colistete, 2010).
12
result, the reputation of the institution was eroded and became vulnerable to pressures
from the outside (de Moura Castro and Verdisco, 1998). All of these dynamics shaped the socalled “Lost-Decade” of the 1980s. Stagnation, increasing income inequality, unbalanced
public accounts, high unemployment, and three and four digits inflation up until 1994
characterized this unfortunate period of Brazilian history (Amman, 2011; Kingstone, 2012;
Kniivilä, 2007).
The SENAI, as well as other vocational training institutions in South America, could not adapt
to the economic downturn and the labour-saving industrial shift which ended up reducing
employment in the formal sector, increasing self-employment and expanding the informal
sector (de Moura Castro and Verdisco, 1998). As the informal sector did not contribute to
financing SENAI, there was no interest in training workers for this sector. Therefore, it might
be argued that the SENAI’s lack of adaptation was directly a consequence of its levy-based
financing scheme.21
The trade liberalization period of the 1990s brought an important reduction of the tariff
levels (Ferreira and Facchini, 2005) and a shift towards the service sector as well as the
middle and high-tech industry sectors, which have been highly subsidized by the government
(Kniivilä, 2007). The structural change increased the demand for workers with higher levels
of qualification (CNI, 2007).22 Contrary to what happened during the “Lost-Decade”, the
upsurge in reputation and enrolment in the SENAI has been progressive. Over the past two
decades, SENAI has targeted actively matching the labour demand of the industry.
Simultaneously, changes in the financing model of the institution (discussed below) played a
role by introducing flexibility to its supply of training. For instance, during the crisis period of
2008-2010, the employment rate of the SENAI graduates did not collapse as it did in the past
(SENAI-DN, 2011).23 In 2011, about 2.5 million people were already enrolled in the SENAI in a
variety of vocational training modalities (SENAI-DN, 2012).24 The performance of the SENAI
graduates throughout the crisis is also possibly a consequence of changes in the selection
process into the training courses. It is likely that over the past two decades the ex-ante
education and family background composition of recent graduates have improved as SENAI
strengthens its reputation. Consequently, the good performance by SENAI graduates might
reflect the ex-ante positive selection into training and not necessarily the true impact of
training on graduates’ labour market outcomes (see also below).
21
In the region, there are some experiences in training the informal sector. The more successful of which was
the Talleres Populares of INA in Costa Rica. Talleres Populares offers open courses in popular crafts and incite
the apprentices to formalize soon after they start their training. These initiatives were never replicated on a
larger scale (CINTEFOR/OIT, 1990 ; De Moura Castro and Verdisco, 1998). The difficulty in training the informal
sector lies on the fact that there is not a single way to proceed, it is costly in term of prestige, and training for
low-skilled occupations is, in most cases, not profitable (De Moura Castro and Verdisco, 1998).
22
During the mid-2000s, more than 85 per cent of new recruits in the oil, machinery, and electronic equipment
sectors had at least a secondary education. (CNI, 2007).
23
According to SENAI-DG (2011), SENAI’s employability rates from 2008 to 2010 were 48 per cent for graduates
from learning and qualification courses and 75 per cent for graduates from technical courses.
24
Vocational training modalities comprehend professional initiation courses, basic industrial apprenticeship
courses, basic vocational qualification courses, mid-level technical courses, post-graduation courses,
graduation courses, training courses and extension courses (SENAI-DG, 2011).
13
1.6.
Financing of the SENAI
The SENAI is principally financed by all industrial companies with a tax of one per cent on all
payrolls that serves as the basis of the contribution to the social security system. The
payment is exempt from federal taxes and is collected by the National Institute of Social
Security (Instituto de Desenvolvimento da Guanabara 1970; Receita Federal, 2008).25
It can be argued that the levy-based resources flowing to the institution are closely linked to
the evolution of real wages and to the growth of the industry, thus having the disadvantage
of being procyclical. As in other training institutions in Latin America, during the so-called
“Lost-Decade”, the budget of the SENAI was negatively affected as the industry moved
towards capital-intensive and natural resource-based activities, though the levy ensured that
structural resources were not cancelled or redirected to other socially productive or
politically profitable activities (de Moura Castro and Verdisco, 1998).
Gasskov (1994) argues that such a financing framework tends to generate a monopoly in the
training market by binding enterprises to the training institution. Consequently, there is no
incentive for employers to provide in-the-job training and the lack of competition in the
training market seems to be a natural outcome of the funding schema. The same study
argues that the financing model tends to standardize technical programmes, offering
reduced opportunities for shop-floor workers. Thus, financing that is fully levy-based can be
inadequate in encouraging further developments to the training system.
During the 1990s, the centralization of the training offer started to reverse with the
implementation of training agreements. In this modality, the SENAI began to supervise
employer funded training rather than provide the training directly (Gasskov, 1994).26
Additionally, revenues associated with the sale of training services to enterprises grew
rapidly, significantly changing the financing structure of the institution. For instance, in 1995
the payroll tax and the sale of services represented 80.8 and 11.4 per cent of the SENAI’s
budget respectively. Five years later, the shares made up by the compulsory contribution
decreased to 71.2 per cent while the sale of services rose up to almost 20 per cent (SENAIDN, 2000).27
The current mixed financing structure has some advantages worth considering. On the one
hand, the levy-based income goes directly to the training corporations, its value is protected
against inflation and it provides a secure and relatively stable funding source (Gasskov,
1994). On the other hand, the progressive specialization in the sale of services offers training
suppliers the possibility to offer ad-hoc training courses at an affordable cost and thus to
compensate the tendency towards the mentioned standardization associated with the levybased financing model. Thus, traditionally training-excluded enterprises have now a greater
number of possibilities to find vocational training courses that are worth it to them. The
flexibility associated with this financing mixture is crucial in the context of an increasing
25
The ordinance nº 4.048/42 indicates that companies with more than 500 employees contribute 1.2 per cent
on all payrolls (SENAI –DR/RJ, 1979; Receita Federal, 2008).
26
See CEDEFOP (1998).
27
At a subnational level, the same trend is observed. For instance, in Guanabara (nowadays Rio de Janeiro) the
share of the compulsory contribution went from 97 percent in 1968 to 60.52% percent in 2013 (Instituto de
Desenvolvimento da Guanabara, 1970; Sistema Firjan, 2013).
14
interdependence between sectors, the incorporation of other agents (universities, technical
schools, consultants) and training modalities such as distance education as well as a labour
market with high levels of informality.
1.7.
Evaluations of the SENAI
Evaluations on the labour market outcomes of vocational training programmes in general
and SENAI graduates in particular, have only recently become more available. As far as we
know, the first attempt to perform a systematic evaluation can be found in Instituto de
Desenvolvimento da Guanabara (1970). This evaluation struggles with a lack of definitions
and agreement about the adequacy of the evaluation criteria. The characteristics of the
sample do not allow making inference about the universe of SENAI graduates in Guanabara
(nowadays Rio de Janeiro). Despite these drawbacks there are some elements worth
considering in this evaluation as they can become relevant to the African context in the
upcoming decades. For instance, Instituto de Desenvolvimento da Guanabara (1970) reports
that a significant amount of SENAI graduates from the period 1965-1968 were working in a
training-unrelated occupation.28 The study indicates that this was due the lack of
employment possibilities and internships vacancies, as well as the compulsory military
service that interrupted the school-to-work transition.29 Moreover, the period 1968-1970
showed a widening gap between the average real wage growth in the industry and the
earnings obtained by SENAI graduates, yet the relative income loss between those aged 1418 and those aged over 18 became smaller during the period 1965-1970 (Instituto de
Desenvolvimento da Guanabara, 1970).
The SENAI-DR/RJ (1979) evaluation depicts the difficulties that SENAI faced in finding
enterprises willing to offer job apprenticeship vacancies. During the 1970s, only a third of all
needed vacancies were available. This report also indicates that SENAI graduates were
positively selected and those who got a job after the training period, in spite of low real
wages, were satisfied with their job situation. De Moura Castro (1979) reports high
estimates of social rates of return, at above 20 per cent of SENAI graduates after secondary
school.
Arriagada and Ziderman (1992) find favourable evidence regarding the returns on vocational
education. They find that trainees and school students obtain comparable earnings when the
former are employed in occupations unrelated to their field of study. Nonetheless, when
trainees are engaged in occupations related to their field of study, their earnings are
significantly higher than those from the formal education system.
Fresneda (2012) finds that graduates from a vocational technical education obtain a
significant wage premium of about 30 percent. At the same time, they would be less
vulnerable to unemployment and informality, and are more likely to enrol in a college
28
The 64 and 47 per cent of workers aged less than 18 and adults respectively, were not working in their field
of training in 1970 (Instituto de Desenvolvimento da Guanabara, 1970).
29
PNAD 2007 shows that S-system graduates (most of them SENAI graduates) who were not working in their
field of specialization in 2007 reached 59 and 43 percent for the same age categories. Due to the fact that it is
impossible to compare both sets of figures, it is difficult to draw conclusions regarding a possible improvement
of the assimilation rates from graduates.
15
education than those who at the most completed a secondary education. This study raises
concerns regarding the selectivity associated with this type of complementary education
since participants in the public system are mostly selected from middle-income households.
Vasconcellos et al. (2010) find a wage premium of 12%, controlling for observables for the
whole sample of individuals with completed secondary education and a return of about 37%
per cent for a restricted sample of children after controlling for parents’ occupation as an
instrument for unobservables.
As a matter of fact, little evidence has been collected to assess the relative performance of
the S-system, including the SENAI, compared to other public and private training institutions.
In 2007, from all the students that enrolled in vocational education at the technical levels
(secondary equivalent), 45% of them did it in institutions run by the public sector, 41%
enrolled in private institutions, and only 14% chose the S-System (including the SENAI).30
2. The data
In this study we indirectly investigate the performance of the SENAI since PNAD 2007 does
not allow disentangling the contribution of the institutions that make the S-system. Based on
a variety of econometric methodologies, we investigate the determinants of participation in
vocational training in 2007, the labour markets outcomes (salaries, hourly wages,
employment, formality, hours of work) of graduates as well as the impact of the training on
workers inter-state mobility (migration). Note that, in most cases, we estimate the impacts
associated with both the S-system and, with other institutions (public and private) that
provide vocational training.
Tables 2 shows the individual and labour market related characteristics by current
enrolment status, type of institution (S-system or others) and type of course (professional
qualification or technical and technological levels) of the those individuals aged 10 or more
that, according to PNAD 2007, could enrol in a vocational training course. Hereinafter, this
group will be called sample 1. Table 3 depicts the same information of those currently
employed.
In order to reduce the underlying heterogeneity of our estimates, the estimation of the
models in the next sections will be based on successively restricted populations. Sample 2
consist of individuals aged between 15 and 29. Sample 3 restricts the sample 2 by
considering only those that live in urban areas. Finally, sample 4 consists of the group of
women in sample 3.31 The estimation based on samples 1-4 provides richer information and
more accurate estimates on the school-to-work transition by reducing the confounding
effects coming from older cohorts, the geographic distribution of institutions supplying
vocational training as well as the systematic differences by gender.
30
According to De Moura Castro (2011), the public sector mainly consists of two modalities of technical
education characterized by abundant regulation under the supervision of the Ministry of Education. These
modalities are secondary technical schools and the “tecnólogos” (two-year post-secondary institutions).
31
Socioeconomic characteristics based on sample 2 to sample 4 are available by the authors upon request.
16
The characterization of the eligible individuals in Tables 2 and 3 shows that enrolment in the
S-system represents only a relatively small fraction of all vocational training offers in Brazil
(14 percent). Nonetheless, amongst workers, the share of S-system graduates increases
significantly (up to 18 percent). This information suggests that in comparison to other
institutions providing vocational training, for employers and employees, the S-system seems
to be an attractive training option.
Both tables also show that there are systematic differences in the socioeconomic
characteristics of trainees that depend on the type of institution (S-system or others) and the
type of received training (professional qualification or technical and technological). In
general, S-system trainees are over proportionally male, non-enrolled in the schooling
system, heads of household, with bi-parental families living in the northern and southern
parts of the country.
Amongst workers, trainees in technological and technical areas earn, on average, more than
their counterparts with a professional qualification, yet trainees in this group, which
represent about 80% of the S-system enrolled workers, seem to earn on average slightly less
than those currently enrolled in other training institutions. Workers’ affiliation to a union is
proportionally more frequent amongst those enrolled in the S-system when compared to
those enrolled in other training institutions. Finally, compared to other institutions, S-system
trainees are over proportionally selected from workers in the production (reparation and
maintenance) of goods and services as well as from the transformation industry (indústria de
transformacão).32
32
This section includes those activities that involve the physical, chemical and biological substances, materials
and components in order to obtain new products. The materials, substances or components transformed are
raw materials produced in the agricultural, forestry, mining, fishery products and other industrial activities.
17
Table 2: Individual and household related characteristics of vocational training eligible
(Sample1).
Nonenrolled
Characteristics for Sample 1
(Aged >10)
Observations
Populations
Enrolment in vocational Training
(%)
Individual Characteristics (% or
years)
Average Age
Enrolled in Vocational Training
320,629
SSystem
1,430
Other
institutions
8,303
Technical and
Technological
SOther
System
institutions
255
1,984
151,223,159
691,690
4,100,807
140,673
993,127
832,363
5,093,934
-
11.7
69.2
2.4
16.8
14.0
86.0
-
Professional Qualification
Total
SSystem
1,685
Other
institutions
10,287
36.5
31.5
25.9
26.8
27.3
30.7
26.2
Female
51.4
41.0
58.2
28.9
53.6
39.0
57.3
Years of education
7.8
10.3
9.8
11.9
11.9
10.5
10.2
Without education
10.6
1.4
1.2
0.0
0.0
1.2
0.9
Incomplete basic
43.9
20.0
30.9
0.0
0.0
16.6
24.9
Complete basic
9.7
14.7
12.0
5.5
6.8
13.1
11.0
Incomplete secondary
6.5
15.0
15.2
20.0
19.5
15.9
16.0
Complete secondary
19.0
39.4
29.1
62.4
57.6
43.3
34.6
Incomplete tertiary
3.7
4.0
3.9
5.6
10.4
4.3
5.2
Complete tertiary
6.6
5.4
7.8
6.5
5.7
5.6
7.4
Children's father- tertiary ed.
1.9
2.3
2.8
1.1
2.9
2.1
2.8
Avg. household education
4.7
5.3
5.3
6.0
6.0
5.4
5.4
Enrolled
Ethnic I (white)
23.5
25.2
46.6
29.0
40.3
25.8
45.4
49.5
52.6
52.8
67.0
59.2
55.0
54.0
Ethnic II (mulatto)
41.8
40.0
39.0
25.0
31.7
37.5
37.6
Ethnic II (others)
8.8
7.5
8.2
8.0
9.0
7.6
8.3
-
29.0
26.0
25.6
29.4
28.5
26.7
73.8
Vocational training in the past
Household related variables
(percent.)
Bi-parental household
74.2
78.3
74.1
77
73
78.1
Household size
4.1
3.9
4.1
3.7
3.9
3.9
4
Head of household
36.0
35.8
19.3
22.6
22.1
33.6
19.9
Spouse
23.9
18.6
17.1
11.3
14.6
17.4
16.6
Children
31.4
38.7
54.6
59.0
56.5
42.2
55.0
Others
8.7
6.9
9.0
7.0
6.8
6.9
8.6
Female with children under 5
6.6
5.3
5.7
4.7
5.2
5.2
5.7
Geographic region (percentages)
North
7.8
6.8
5.9
4.8
5.6
6.4
5.8
Northeast
27.5
20.2
22.3
8.8
16.9
18.3
21.2
Southeast
42.9
43.4
49.2
59.1
53.7
46.1
50.1
South
14.6
21.4
15.6
19.9
19.3
21.2
16.3
Midwest
7.3
8.2
7.1
7.3
4.5
8.0
6.6
Source: own calculations based on PNAD 2007.
18
Table 3: Labour market related characteristics of vocational training eligible (employed).
Never
enrolled
Characteristics for Sample 1
(Aged >10)
-
Enrolled in Vocational education
Population (with labour earnings)
Total
160,353
911
Other
institutions
3,718
1,080
Other
institutions
4,894
75,623,882
444,438
1,854,157
94,779
595,005
539,217
2,449,162
14.9
62.0
3.2
19.9
18.0
82.0
S-System
Observations
Technical and
Technological
Other
S-System
institutions
169
1,176
Enrolment in voc. Training (%)
S-System
Labour markets statistics
Labour earnings (Reais 2007)
955
929
953
1,021
960
945
955
Working hours (weekly)
39.5
39.7
37.1
40.5
38.5
39.8
37.4
Union membership (%)
16.9
19.3
16.6
19.3
18.2
19.3
16.9
Formality (%)
32.2
47.4
29.9
56
45.7
49
33.1
Recent migrant (%)
7.7
8.7
8.3
8.6
9.1
8.7
8.5
C.E.Os and Managers
4.7
5.1
4.9
4.2
4.9
4.9
4.9
Professionals
6.5
4.6
9.6
3.6
8.4
4.4
9.3
Technicians
7.2
9.4
12.4
16.1
21.4
10.5
14.5
Administrative workers
8.5
11.1
14.0
14.6
20.9
11.7
15.6
Service workers
20.7
18.4
22.9
10.9
14.9
17.1
21.0
Tradesman
10.0
10.9
12.4
10.1
10.5
10.8
11.9
Agriculture workers
18.8
5.9
5.1
2.5
2.6
5.3
4.5
Goods, services producers
23.1
33.2
18.0
36.9
15.5
33.8
17.5
Military
0.6
1.6
0.7
1.1
0.8
1.5
0.7
Others
0.0
0.0
0.1
0.0
0.1
0.0
0.1
Agriculture
18.8
6.2
5.3
2.5
2.6
5.6
4.6
Other industrial activities
0.8
0.7
1.0
2.9
2.3
1.1
1.3
Transformation industry
14.1
23.0
16.1
36.2
18.6
25.3
16.7
Occupations (percentage)
Sectors (percentage)
Construction
6.9
6.1
3.4
2.5
3.2
5.5
3.4
Commerce
17.8
22.6
20.7
24.0
20.8
22.9
20.7
3.9
4.3
4.7
2.5
2.1
4.0
4.1
4.7
6.1
3.5
1.8
4.6
5.4
3.7
4.8
5.6
6.1
3.4
7.3
5.2
6.4
Education, health and social serv.
8.8
6.8
15.1
11.0
19.9
7.5
16.2
Domestic services
8.1
4.2
7.1
1.5
4.0
3.8
6.3
Other collective and social serv.
3.9
6.5
7.2
3.3
4.5
5.9
6.6
Other activities
7.5
7.9
10.0
8.4
10.1
7.9
10.0
Hotels and Restaurants
Transport and
telecommunication
Public administration
Source: own calculations based on PNAD 2007.
Note: Formality is defined as individuals contributing to social security system, contract, public servants, military and employees with
more than five workers in non-agricultural activities.
19
3. Determinants of participation in vocational training
3.1.
Model of participation
In order to avoid endogeneity issues, we consider only those individuals eligible for
vocational training without a migration background. This is only a fraction of the population,
representing approximately 55 percent of those eligible individuals described in Table 2.
Based on sample 1 (all individuals older than 10) to sample 3 (individuals aged between 15
and 29 in urban areas), we estimate survey design-adjusted probit models for the probability
of current participation in a vocational training course. In order to explain the enrolment
decision, we include the age of the potential participant, the individual’s level of education,
gender, family education indicators to control for unobservables and the ethnic/racial
backgrounds of potential participants.33 Additionally, the model includes the individual’s
position within the household (head of household, spouse, children and others), the biparental condition of the family, the interaction between females and children under the
age of five, the household size, region dummies, as well as the current enrolment in the
schooling system and the employment status. Furthermore, we include interactions
originating in the labour market for the employed individuals. The interactions include:
currently working in the formal sector, union membership, hours worked during the past
month as well as controls for the economic sector and occupation. 34 Table 4 shows the
results of the probability estimation by sample (1-3) and type of institution. Further analysis
can be carried out based on Tables A.1 and A.2, which show probability models for the
enrolment in training courses of professional qualification and in courses at the technical
and technological level respectively.
Results in Table 4 indicate that, in general, trainees tend to be younger and the ones in the
S-system tend to be older than their counterparts in other training institutions. This age
effect is stronger for young cohorts (sample 1 versus sample 2) while the rural condition
seems to be uncorrelated to the age-enrolment profiles (sample 2 versus sample 3).
Regarding the selection in education, our results suggest an inverse U-shaped relationship
between the educational level and the enrolment probability. The inflection point is found at
the level of completed secondary education. Those in school-to-work transition age are
additionally positively selected on the human capital endowments of their families (average
household education). Consequently, the current patterns of enrolment in vocational
training confirm the concerns that vocational education does not particularly attend to the
needs of less skilled populations, and therefore, the SENAI and other vocational training
institutions could do more enhance the social mobility of the disadvantaged (see Fresneda,
2012).
33
The controls for unobservables consist of an index of tertiary educated father for children and the average
household education, excluding the potential participant.
34
Formal workers are defined as those contributing to the social security system in their principal or secondary
occupation, those with formal working contracts, military, public servants, and employers with more than five
employees in non-agricultural activities.
20
Table 4: Probability model for the enrolment in vocational training, 2007
Survey: Probit regression
Population
Variables / Type of institution
Age
Incomplete basic
Complete basic
Complete secondary
Incomplete tertiary
Complete tertiary
Children x father's educ. (T)
Female
Ethnic I (white)
Ethnic II (mulatto)
Spouse
Children
Other household member
Household size
Female x children under 5
Northeast
Southeast
South
Midwest
Enrolled in school system
Employed
Interactions with employed
Formal
Union membership
Hours of work
Industrial activities
Manufacturing
Education and health
Other Services
Managers
Professionals
Technicians
Craftsman
Service workers
Trade workers
Goods, services producers
Constant
Observations
Population
Design df
F-statistic
p-value
Number of Strata
Number of PSU
Dependent variable: enrolment in vocational training, by type of institution
Sample 1
Sample 2
Sample 3
Other
Other
Other
S-system
S-system
S-system
institutions
institutions
institutions
-0.007***
-0.013***
-0.029***
-0.032***
-0.030***
-0.032***
0.260**
0.369***
0.200
0.368***
0.141
0.416***
0.747***
0.746***
0.501**
0.717***
0.452**
0.730***
0.911***
0.921***
0.777***
0.903***
0.716***
0.891***
0.940***
1.036***
0.861***
1.102***
0.799***
1.055***
0.490***
0.569***
0.353*
0.594***
0.279
0.594***
0.611***
0.802***
0.331
0.873***
0.272
0.837***
-0.320***
-0.110**
-0.397***
-0.196***
-0.399***
-0.198***
0.016**
0.012***
0.022**
0.006
0.008
-0.002
-0.203***
0.073***
-0.308***
0.029
-0.324***
0.020
-0.035
-0.037
-0.066
-0.036
-0.052
-0.043
-0.003
-0.024
0.010
0.003
0.018
0.010
0.096**
0.024
0.132***
0.045*
0.143***
0.061**
-0.085
-0.051*
-0.140
-0.050
-0.115
-0.055
0.050
0.077***
-0.028
0.054
0.005
0.066
-0.028
0.021
-0.120
-0.011
-0.088
-0.017
-0.035***
-0.037***
-0.048***
-0.034***
-0.044***
-0.030***
0.058
-0.099***
0.083
-0.115**
0.020
-0.112**
-0.050
0.053
0.005
0.047
0.012
0.058
0.041
0.207***
0.122
0.168***
0.103
0.166***
0.163
0.232***
0.184*
0.175***
0.182
0.170***
0.055
0.075
0.106
0.048
0.038
0.016
0.060
0.288***
0.043
0.245***
0.020
0.215***
0.137
-0.293***
0.111
-0.477***
0.071
-0.575***
0.133***
0.037
-0.002***
0.033
0.107**
-0.025
0.061
0.172**
0.078
0.027
0.168
0.120
0.196**
0.243**
0.275***
-2.985***
174872
83282219
5455
18.276
0.000
637
6092
-0.081***
0.117***
-0.001***
0.309***
0.129***
0.078*
0.154***
0.150***
0.387***
0.441***
0.538***
0.423***
0.459***
0.433***
0.318***
-2.318***
180305
85990337
5493
82.743
0.000
699
6192
0.224***
0.072
-0.003***
-0.003
0.068
-0.105
-0.070
0.008
0.180
0.248
0.431**
0.289*
0.339**
0.484***
0.360**
-2.316***
67046
31161384
4999
10.278
0.000
627
5626
-0.073**
0.147***
-0.001***
0.555***
0.114***
0.039
0.130***
0.117*
0.403***
0.601***
0.652***
0.593***
0.610***
0.570***
0.416***
-1.810***
70351
32768719
5040
40.617
0.000
649
5689
0.245***
0.071
-0.004***
-0.190
0.064
-0.096
-0.043
0.018
0.264
0.323
0.514*
0.369
0.413
0.549**
0.425
-2.137***
54798
25145643
4658
8.717
0.000
592
5250
-0.066*
0.159***
-0.001***
0.566***
0.111***
0.012
0.115**
0.131**
0.495***
0.688***
0.730***
0.671***
0.660***
0.630***
0.466***
-1.738***
57747
26567877
4691
27.284
0.000
592
5283
Excluded Categories: No education, Male, other ethnic groups, heads of household and northern region, non-enrolled, unemployed.
Note: Samples (1) consist of individuals aged + 10. Sample (2) is (1) aged between 15 and 29. Sample (3) is sample (2) in urban areas. All
treatment effects are significant at least at the 5% level.* Not significant at the 10% level.
Source: Author's calculation based on PNAD 2007 data. * p<.10, ** p<.05, *** p<.01
Confirming the figures in Table 2, enrolment in the S-system is strongly biased against
women while enrolment in other institutions favours women in older cohorts (sample 1
versus sample 2). The ethnic dimension does not appear to play any role in determining
enrolment. Consequently, it is possible to state that differences in Table 2 respond
exclusively to different endowments and different race groups.
Those belonging to a bi-parental household are more likely to be enrolled in vocational
training; however, the bi-parental condition of a household also favours the enrolment in
the S-system more. Consequently, if vocational education is considered an investment in
human capital, it is possible to conclude that investments associated with the S-system are
higher than those in other institutions. Interestingly, we find that women with children
under the age of five reduce their probability of enrolling in other institutions, while this
21
probability is not affected in the case of enrolment in the S-system. This result is not trivial. It
implies that the cost of child rearing (associated with enrolment in the S-system) is spread
across women of all ages and it is not attached exclusively to woman at the reproductive
age. At this point, it would be recommendable to further study the enrolment mechanisms
of the S-system that avoid women with responsibilities, causing them to become
disadvantaged when compared to women without small children. At the same time, the
general gender bias in the S-system should be called into question.
Regarding the geographic determinant of enrolment, significant coefficients by regions can
be interpreted as regional distortions in the supply of vocational education relative to the
mass of workers. Our results suggest that the S-system tends to, in some degree, oversupply the population in the south of the country (Paraná, Santa Catarina and Rio Grande do
Sul). The other institutions providing vocational training over-supply to a greater extent in
the same regions, but also do so in the south eastern region (Espírito Santo, Minas Gerais,
Río de Janeiro and São Paulo).
While enrolment in the school system does not affect the probability of being trained in the
S-system, it strongly correlated to training activities in other institutions. It seems that there
is a neutral relationship between formal education and the S-system and a complementary
relationship with the other institutions. This asymmetry could reflect the fact that being
enrolled in the S-system is associated with a balanced mixture of concomitant and
subsequent provisions of training while the enrolment in other institutions has mixture
biased towards integrated and concomitant modalities. The same idea is reinforced by the
fact that trainees in other institutions are less likely to currently have a job than their nontraining-enrolled counterparts, however the likelihood of enrolment in the S-system is not
associated with the employment status. These results suggest a higher relative prevalence
of training of currently employed in the S-system compared with other suppliers of
vocational education.
Amongst the employed population, workers in the formal sector are more likely to enrol in
the S-system, while their counterparts in the informal sectors seem “to prefer” other
training institutions.35 While workers in the service and trade sector as well as in the
production (reparation and maintenance) of goods and services have a higher probability to
enrol in the S-system, agricultural workers are clearly less likely to enrol in other (private and
public) training institutions.
3.2.
The selectivity process into the S-system and other
institutions
In courses of vocational education, there are systematic selectivity differences between
those enrolled in upper level courses (technical and technological - secondary and tertiary
equivalent) and those enrolled in professional qualification courses (middle-low qualification
level).
35
The high correlation between union membership and formality may explain why union status appears to be
uncorrelated with training in the S-system.
22
Based on results in Tables 4, we find that for upper level courses, women, children and other
household members as well as those currently enrolled in the schooling system are less
likely to be selected into the S-system when compared to other alternatives of vocational
training. Note that the bias against female enrolment in the S-system is important in
magnitude, highly significant, and is found even after controlling for employment status and
its interactions with economic sectors and occupations, the position within the household as
well as having children under their responsibility. On the contrary, males currently employed
in formal jobs as well as those coming from bi-parental households have a higher relative
likelihood of being enrolled in the SENAI, while union status plays no role. Older cohorts,
technicians, those working in other industrial activities and those living in the southeast of
the country or with a tertiary education are relatively less likely to be enrolled in the Ssystem. On the contrary, mulattos are positively selected into the S-system when compared
to other alternatives. Amongst younger cohorts (aged between 15 and 19 – sample 2) we
find that the individual’s position within the household plays a key role. It seems that
enrolment in the S-system is more likely amongst heads of household (excluded category) as
well as amongst those in full time jobs (working more hours a week).
Regarding the courses of professional qualification (middle-low skills level), compared with
other alternative vocational training institutions, those with a higher education, or who are
children (lower domestic responsibilities), or have smaller families, or are employed in
formal jobs or in the manufacturing sector are more likely to be selected into the S-system.
Contrarily, amongst young cohorts, while union status, the ethnic background, enrolment,
and the employment status play no role, women and those currently in the school are less
likely to attend the S-system.
We hypothesise that these cohort effects can be understood as an indication those with
domestic responsibilities expect to improve their labour market outcomes by enrolling Ssystem upper level courses. On the contrary, those with less domestic responsibilities tend
to prefer professional qualification from institutions outside of the S-system. This is
presumably because of the flexibility associated with the heterogeneity of the many
institutions, which according to PNAD 2007 provide 86 per cent of the total training in Brazil.
In order to investigate this hypothesis, the next section focuses on investigate how training
impacts labour markets outputs.
4. Vocational training graduates and labour market
outcomes
In this section, we investigate whether past vocational training episodes influences current
labour market outcomes. In particular, we are interested in assessing its impact on monthly
labour earnings, monthly hours of work, hourly wages (as an indicator of productivity) and
the probability of being employed in the formal sector (as defined previously). All of these
average treatment effects estimations are based on valid observations of employed
individuals that reported labour earnings where the reported estimates are extendable to
the whole working population and not only to the treated (trained) population. Most of
these individuals have never been enrolled in any vocational training and are consequently
considered as a control group. Next, based on the eligible population for vocational training,
23
we estimate the impact of vocational training on the employment probability (also by
institutions). Finally, by restricting the sample to graduates, we estimate the differential
impact of S-system training on the probability of working in the same training area. All
models are estimated across sample 1 to 4, as well as by individual’s position within the
household. The idea of such a disaggregation strategy is to account for the
multidimensionality of vocational training. Thus, our results will reflect the differences in
impact by cohort, geography, gender and individual’s household position.
4.1.
Impact evaluation of SENAI
In order to estimate the average treatment effect (ATE) on the labour market outcomes
caused by a previous training experience, we rely on inverse probability weighting (IPW)
estimators.36 The methodology was chosen taking into consideration the dynamic nature of
the enrolment decision. Vikström (2014) suggests that this estimator is suitable and even
preferred to other matching estimators in treatment evaluations in a discrete time setting
where the intervention could begin at any point in time in the past.
Consider outcome
(for example, labour earnings) and a binary treatment variable
{ } for instance, the graduation from a vocational training course provided, for
example, by the S-system (or alternatively by other institution). Additionally, assume that
the potential outcomes
and depends on the treatment (training). Then, it is necessary
to obtain the mean labour earnings for trained individuals
which is observed
37
when
and unobserved when
. The treatment effect estimation problem is thus
not different than solving a missing-data problem.
An IPW estimator for
is :
∑
⁄
where
is the probability that
(to be trained or graduate), which is a function of
the observable determinants
. The IPW estimator weighs asymmetrically on those
observed outcomes event though the observation was not probable.
In order to properly estimate the effect of S-system training as well as the effect of training
by other institutions, a probit model is estimated where the dependent variable takes the
value of one if the individual was trained sometime in the past and zero if he/she did not
have any training experience. The vector
of determinants contains only exogenous
variables. That is, only variables that cannot be affected by the current outcome (for
example, labour earnings), as well as variables that may have changed between the training
event and the current outcome (for example, to get married). Thus, the selection of variables
36
IPW dates back to Horvitz and Thompson (1952). Since then, researchers have been actively trying to extend
this approach to deal with modern treatment-effect estimation problems in the field of biostatistics and
econometrics (see Robins and Rotnitzky, 1995; Robins, Rotnitzky, and Zhao, 1994 and 1995; Wooldridge 2002,
2007 and 2010; Hirano, Imbens, and Ridder, 2003).
37
The observed data (
corresponds to
for trained individuals and is missing for their non-trained
counterparts.
24
determining the probability of enrolment is difficult since the temporal dimension of the
training event is missing and important information prior to the training event is also
neglected. Additionally, propensity score methodologies rely on the assumption that
unobservables play no or at least only a small role in determining participation (see Dehejia,
2005; Zhao, 2004). For this reason, the evaluation literature is abundant in
recommendations about the inclusion of instruments for talent, motivation and other nonobservable characteristics. Nevertheless, even when the availability of instruments is
reduced, at least for children, we instrumented those unobservables by using the education
achievements of their parents.
Table 5: Average treatment effect of vocational training on selected variables, 2007
Vocational training (All courses)
Monthly earnings
Monthly working
hours
Ssystem
Ssystem
Other
inst.
Other
inst.
Hourly labour
earnings
Ssystem
Other
inst.
Formality
Ssystem
Other
inst.
Employment
Employment
in same area
Ssystem
Other
inst.
S-system
Absolute effect of training (units)
Sample 1
68
109
0.4
1.7
0.30
0.60
0.09
0.08
0.10
0.10
0.08
Sample 2
168
63
-1.1
2.0
0.86
0.35
0.09
0.09
0.07
0.09
0.11
Sample 3
87
54
-0.4
1.6
0.26
0.25
0.06
0.08
0.07
0.06
0.12
Sample 4
0
16
-0.2
2.0
-0.23
-0.02
-0.03
0.04
0.09
0.09
0.12
Predicted outcome mean without training (units)
Sample 1
859
892
180
180
5.50
5.73
0.57
0.59
0.56
0.57
0.50
Sample 2
593
607
182
182
3.74
3.85
0.54
0.56
0.58
0.59
0.37
Sample 3
637
645
183
183
4.03
4.12
0.58
0.60
0.62
0.63
0.38
Sample 4
595
600
170
171
4.24
4.26
0.59
0.61
0.49
0.51
0.35
Relative effect (percentage)
Sample 1
7.9
12.2
0.2
0.9
5.5
10.5
16.0
13.2
17.3
18.1
15.9
Sample 2
28.3
10.4
-0.6
1.1
23.1
9.0
16.1
16.6
12.3
14.7
30.8
Sample 3
13.7
8.4
-0.2
0.9
6.5
6.0
9.9
13.4
10.5
10.0
32.0
Sample 4
0.0*
2.7
-0.1
1.2
-5.5
-0.5
6.0
6.0
17.6
17.6
33.2
Source: Authors calculations based on PNAD 2007.
Note: Samples (1) consist of individuals aged + 10. Sample (2) is (1) aged between 15 and 29. Sample (3) is sample (2) in urban areas. Sample (4)
corresponds to women of (3). All treatment effects are significant at least at the 5% level.* Not significant at the 10% level.
Based on the PNAD 2007 data, the vector consists of the age and its squared, parental
education (in the case of those in a child’s position within the household, gender, ethnic
background (white, mulatto or others), rural area, federative unit and educational level. To
avoid endogeneity issues and make our estimation more reliable, we exclude currently
enrolled individuals (school or training) as well as those with an inter-municipal migration
background since they could move after training and therefore biasing the estimated
impacts.
Table 5, 6 and 7 refer to the impacts of all vocational training, professional qualification
courses and technical/technological courses, respectively. The tables show the IPW
estimated impacts on all variables of interest by institution and sample. A disaggregated
view (by individual’s position within the household) of the impacts can be found in the
Appendix Tables A.3 to A.8.
25
Table 5 shows the significant labour earnings premium obtained by S-system graduates,
which on average was about of 28 percent for workers aged between 15 and 29 (sample 2).
The fact that the impact decreases in sample 3 (only urban areas for the same cohort)
indicates that the S-system has an even greater premium in rural areas. The relatively low
premium in sample 1 indicates that older cohorts do not obtain a significant labour earnings
premium and even probably suffer income losses. This might either relate to lower effects
of SENAI in the past or that these effects dissipate over time; as we are using cross-sectional
data, we are unable to differentiate among the two hypotheses. The contrary occurs with
other suppliers of training. They exhibit higher returns for older cohorts and a small
difference between rural and urban areas. In both cases, women do not obtain an important
training premium (sample 4). The impacts on the intensity of work are very small. Instead,
training appears to increase labour productivity and as a consequence, hourly wages rise.
This is also true for the monthly earnings patterns with the exception of urban women
(sample 4). For this group, training seems to negatively affect its labour productivity.
In general, vocational training induces higher levels of formality with no big difference
existing between the S-system and other institutions. The same happens regarding
improvements in the employability of training graduates. The most important differences
between the S-system and other training institutions are related to the specificity of the
training. As a matter of fact, the young cohort of S-system graduates are about 30 percent
more likely to work in an occupation related to the training area.
Regarding differences between professional qualifications and upper level training (Table 6
and 7), in the case of young women in urban areas (sample 4) the general loss of
productivity can be decomposed into a small but widespread decay in productivity and work
intensity in the case of professional qualifications. Contrary to this, women with upper level
training increase their labour earnings by increasing the number of hours of work, even
when their hourly salaries decline by about five percent. Note that the positive impact of the
S-system on labour earnings is much more accentuated for those with upper level training.
26
Table 6: Average treatment effect of professional qualification training on selected variables,
2007
Monthly earnings
Monthly working
hours
Hourly labour
earnings
Ssystem
Other
inst.
Ssystem
Other
inst.
Sample 1
68
117
0.6
2.4
0.34
0.59
0.09
Sample 2
157
52
-1.1
2.9
0.84
0.24
Sample 3
84
40
-1.3
2.5
0.28
Sample 4
-17
16
-2.3
2.7
-0.22
Sample 1
856
864
180
180
5.47
5.55
0.57
Sample 2
591
596
182
182
3.73
3.77
Sample 3
634
634
183
183
4.03
Sample 4
595
591
170
170
4.23
Sample 1
8.0
13.6
0.6
2.4
6.3
10.6
Sample 2
26.5
8.7
-1.1
2.9
22.4
Sample 3
13.2
6.3
-1.3
2.5
Sample 4
-2.9
2.6
-2.3
2.7
Ssystem
Other
inst.
Formality
Ssystem
Other
inst.
Employment
Employment
in same area
Ssystem
Other
inst.
S-system
0.08
0.10
0.10
0.07
0.09
0.10
0.07
0.09
0.12
0.11
0.06
0.08
0.07
0.06
0.13
-0.01
-0.05
0.03
0.10
0.09
0.13
0.58
0.56
0.57
0.51
0.54
0.55
0.58
0.59
0.35
4.04
0.58
0.59
0.62
0.62
0.36
4.20
0.59
0.60
0.49
0.50
0.33
16.2
13.4
17.3
17.3
13.4
6.3
16.1
17.4
12.3
14.6
34.8
7.1
2.7
9.6
13.9
11.2
9.4
36.5
-5.2
-0.3
-8.1
4.4
20.4
17.1
40.0
corresponds to women of (3). All treatment effects are significant at least at the 5% level.
Regarding the impacts on formality and employment, the differences are concentrated in
the dissimilar impact on women. On the one hand, S-system graduates from courses of
professional qualifications are more likely to have a job as well as a job in the same area of
training. The price for that is that they are also less likely to work in the formal sector. On the
other hand, S-system graduates from upper level courses are more likely to formalise, but
with less success in terms of employability and permanence in the training area.
This information suggests that, while women with professional qualification training seem to
stagnate in low prestige occupations, their counterparts who received upper level training
(technical and technological) seem to enjoy higher mobility levels.
27
Table 7: Average treatment effect of technical and technological training on selected
variables, 2007
Technical and Technological level
Monthly earnings
Monthly working
hours
Ssystem
Ssystem
Other
inst.
Other
inst.
Hourly labour
earnings
Ssystem
Other
inst.
Formality
Ssystem
Other
inst.
Employment
Employment
in same area
Ssystem
Other
inst.
S-system
Sample 1
65
86
-1.3
-0.3
0.02
0.64
0.08
0.08
0.10
0.12
0.12
Sample 2
235
105
-0.8
-1.2
1.01
0.77
0.09
0.08
0.07
0.09
0.09
Sample 3
104
106
4.6
-1.5
0.14
0.75
0.07
0.07
0.04
0.08
0.09
Sample 4
106
18
12.7
-0.3
-0.34
-0.05
0.09
0.07
0.02
0.10
-0.05
Sample 1
880
970
181
180
5.66
6.24
0.58
0.62
0.57
0.58
0.53
Sample 2
605
649
182
182
3.79
4.14
0.55
0.60
0.59
0.61
0.48
Sample 3
654
686
183
182
4.08
4.40
0.59
0.63
0.63
0.64
0.49
Sample 4
595
631
170
171
4.28
4.47
0.60
0.64
0.60
0.53
0.46
Sample 1
7.4
8.9
-0.7
-0.2
0.3
10.2
14.4
12.5
16.9
20.0
23.2
Sample 2
38.8
16.1
-0.4
-0.7
26.6
18.5
16.5
13.7
12.2
15.5
18.1
Sample 3
15.9
15.4
2.5
-0.8
3.4
17.1
11.4
11.6
7.0
12.1
18.2
Sample 4
17.8
2.8
7.5
-0.2
-7.8
-1.0
14.1
11.2
3.1
19.0
-11.3
corresponds to women of (3). All treatment effects are significant at least at the 5% level.
4.2.
The Heckit approach
One of the limitations of the propensity score methods when estimating average treatment
effects lies on its inability to control for factors beyond the researcher’s control.
Unobservables are then a potentially dangerous source of bias. These methods offer great
flexibility because they do not rely too much on a functional forms or impose strong
statistical assumptions on the outcome equation. On the contrary, the Heckman sample
selection model (Heckit approach) has complementary characteristics that can be used to
investigate the extent to which unobservables are associated with the outcome variable. The
disadvantage is that it imposes strong restrictions to the joint behaviour of unobservables.
The Heckit model is thus estimated in this evaluation context to assess the impact of
vocational training provided by S-system on labour earnings after controlling for
unobservables.
In order to avoid endogeneity issues, we consider only non-enrolled individuals without an
inter-municipal migration background. The included variables in the outcome equation are
the level of education, S-system background, the average level of education of the family,
potential experience and its squared, union membership, the number of children in the
household, ethnic background (white, mulatto or others), rural condition as well as controls
for economic sector and occupation. The selection equation relies on almost the same
28
determinants, the exclusion variable being a dummy variable whether an individual lives in a
bi-parental household. Note that because the inverse Mill’s ratio (IMR) is a nonlinear
function of the explanatory variables in the probit model (first-step), then the second-stage
equation is identified even if there is perfect overlap in the variables in the selection and
outcome equations. Having a variable in the selection equation that it is not included in the
outcome equation is, however, recommended since this is parsimonious procedure.
Table A.9 in the appendix presents the results of the wage corrected equation for samples 14. In general, this model confirms the positive impact of S-system training on earnings and
employment. The fact that the young cohort at the country level shows a negative
coefficient for the IMR suggests that there are unobserved factors upwardly biasing the
returns to endowments. Thus, there is the potential that unobservables may bias upward
the estimations based on the propensity score in at the very least sample 2. Consequently, it
is safe to conclude that the S-system premium on monthly labour earnings is located
between almost 9 and 30 percent, being consistent with the findings by Fresneda (2012) and
Vasconcellos et al. (2010). Additionally, we find a significant impact on the employment
prospects of young individuals. The selection equation confirms that the S-system increases
the chances of employment for women. This positive outcome is occurring in the context of
extremely low enrolment chances for women relative to men. Consequently, the questions
whether women’s relative difficulties in accessing the S-system are basically explained by
demand factors (women’s preferences) or by discrimination, are crucial.
4.3.
Rural urban and gender gaps
Given the heterogeneity in the impacts of the S-system across geographic areas (rural-urban
as the differential between estimates based on samples 1 and 2) as well as across gender
(samples 3 and 4) there are enough reasons to believe that the impacts of S-system training
on the labour income distribution could be disequalizing since the earnings premium
appears to configure a sloping ladder. The subsidiary hypothesis is that the S-system
promotes, in this order, young heads of household, children, and then (presumably after
other household members) women. For example, Table A.3 in the Appendix shows that
heads of household aged between 15 and 29 trained in the S-system (in all levels) obtain an
average premium in 2007 reais of 110 per month in urban areas (14.9%). Children do obtain
a small return of almost 24 reais per month (3.8%) while women seem on average, to obtain
no premium at all.
Since the returns to the accumulation of human capital are not homogeneously distributed
amongst groups (gender and urban-rural areas), labour income inequality levels depend on
the earning-gap dynamics between these dimensions. In order to investigate this issue, we
use quantile regressions to assess the size of the S-system premium across labour income
distributions by gender and by urban/rural area.
Figure 1 shows the degree to which the S-system training premium increases the gender
labour earnings gap. On the one hand, the premium for men is higher than for women. On
the other hand, the premium for women is flat (about 6%) and starting in the last quartile
increases, reaching about 15% while the male premium is flat until the second quintile and
29
then grows reaching about 25% at the upper tail of the distribution. The conclusion is that
the S-system contributes to the widening of the labour income distribution trough its
enhancing impact on the gender gap.
Contrary to this, the S-system premium contributes to closing the rural-urban labour
earnings gap. Figure 2 shows that across almost the whole income distributions, that is,
quantile by quantile, there is an extra premium for S-system graduates working in rural
areas. It can be argued that this premium is associated with the lack in the supply of skills in
low-density areas. This information is also confirmed in Table 5 by comparing the impact of
the S-system on labour earnings in samples 2 and 3 (sample 2 considers young individuals at
the country level, while sample 3 restricts the computation of the effect to only urban
areas). The fact that the impact in sample 3 is smaller across household members than in
sample 2 means that the S-system premium in rural areas is larger than the one in urban
areas. Thus, regarding the contribution of the S-system to labour income inequality changes,
the geographic and gender dimensions work in opposite directions causing the impact of the
S-system on labour income inequality is rather ambiguous.
30
0.00 0.05 0.10 0.15 0.20
0
.2
.4
.6
.8
1
0
.2
.4
.6
.8
1
0.40
0.20
0.00
Rural areas
0.60
Urban areas
Figure 1: S-system graduates’ premium across quantiles of labour earnings, by gender in
2007 (Source: Authors calculations based on PNAD 2007)
Figure 2: S-system graduates’ premium across quantiles of labour earnings, by areas in 2007
(Source: Authors calculations based on PNAD 2007)
31
5. Mobility and the S-system
The possibility that vocational training could discourage social mobility is one of the most
recurrent arguments against the expansion of this type of education (see Goldthorpe and
Erikson, 1992; Müller and Pollak, 2005; Shavit and Müller, 1998; Breen and Jonsson, 2007;
Müller and Gangl, 2003; Buchmann and Park, 2009). It is argued that vocational training
discourages the continuation of studies and therefore contributes to increasing the degree
of stratification of the society, however it could also have positive effects on geographical
mobility. There is a gap in the literature regarding this unexplored dimension that associates
internal migration with vocational training and hence the analysis of social mobility is
incomplete if its geographical dimension is missing. In order to address this lack of
knowledge, we implement a methodology to see whether graduates of the S-system were
more likely to migrate to other states besides their non-trained counterparts. If this is so,
losses coming from the occupational segmentation could be partially compensated through
gains in geographic mobility.
Empirically, we adopt a similar approach to the one used by Brücker and Trübswetter (2007)
in studying the selectivity process associated with the migration of workers between East
and West Germany after reunification and the role of wage differentials. Villalobos Barría
(2012) similarly investigates the factors constraining internal migration in Chile, while
Guilietti et al. (2012) tries to explain the determinants of the self-employment decision of
rural-urban migrants in China.
The approach is based on Borjas (1987) who uses the Roy (1951) model to derive a selection
model of migration that accounts for unobservables. The main feature of the model is that
there is no relationship between the selection process generated by unobserved and
observed characteristics.38
Borjas (1988) presents a model with random mobility costs.39 In this model, the wage at the
destination
and the wage at origin
are joint-normally distributed and the
migration costs are not fixed but rather a proportion to home income. Migration occurs if
the wage differential is greater than the migration cost (
), however wages
at the origin and destination cannot be observed simultaneously and comparing mean wages
yields endogeneity bias. In order to overcome this missing data problem, we employ a threestep strategy to obtain consistent and efficient estimates of the individual probability to
migrate and its determinants.40
In the first step, a probit reduced-form model for migration serves as the starting point for
the estimation, where selectivity corrected wage equations (second step) for migrants and
stayers account for the role of unobservable. Finally, in the third step, a structural
38
The standard Roy model does not consider any switching costs. As a consequence of this, important
information is not taken into account if the costs of moving are inversely related to the amounts of human
capital (workers’ heterogeneity). See Villalobos Barría (2012) and Klasen et al. (2014).
39
This model is a generalization of the Roy model that relaxes the assumption of constant moving costs by
allowing correlation between non-observed abilities and moving costs.
40
In this study, we avoid the underestimation of the coefficient’s standard deviations by accounting for the
complex survey design of the PNAD 2007 household survey.
32
probability model, which properly accounts for earning differentials, explores the structural
determinants of migration distinguishing their influence on migration by working through
earnings differentials and through other mechanism uncorrelated to it.
Using this framework, it is possible to assess the relationship between the S-system and the
wage differential as well as the likelihood that such a vocational training encourages interstate migration, thus enhancing the mobility prospects of its beneficiaries. Villalobos Barría
(2012) presents a detailed description of the estimation procedure, which relies on a
switching regression model by Goldfeld and Quandt (1973) with endogenous switching
(Maddala and Nelson, 1975; Maddala, 1983).
The migration index function in the first step is estimated using the probit Maximum
Likelihood (ML) estimator. In our definition of migration, inter-state movers are individuals
who moved anytime within the last four years from one state to another state. The model
includes as explanatory variables the level of education of the potential migrant, the
potential experience and its square, gender, the number of children older than five in the
household (born before migration time span), indigenous status (as above), and controls for
the regions at origin. All variables in the reduced form are expected to proxy the
characteristics of interest at origin (ex-ante). Based on PNAD 2003, state-level
unemployment rates at origin are also included to reflect the hypothesis that unemployment
encourages migration outflows. The log of the population at origin aims to control for the
availability of public goods.
In the second step, an unbiased and non-endogenous labour earnings differential is
calculated and in the third step is included (in logs) together with its square (to allow for
nonlinearities) to the structural estimation of the migration probabilities. As explanatory
variables, besides the potential wage differential, the structural equation includes the state
unemployment rates at origin, the log of the population at origin and the age of the
potential migrant as the migration propensity decreases with age (Greenwood 1993). As
discussed earlier, of interest is the inclusion of a variable indicating if someone received
training in the S-system. Finally, controls for the number of children in the household, the
ethnic origin (white, mulatto or others) as well as dummies for the states of origin are
considered.
The estimation is based on employed individuals aged between 30 and 40 (age bracket
adjacent to the one used to study the school-to-work transition), neither enrolled in any
schooling nor training activity, who lived in the country since at least 2002, and without
children under the age of five. The age restrictions as well as the non-enrolment conditions
are expected to reduce the possibility that someone could migrate in order to enhance the
available educational chances as well as to avoid the confounding non-causal correlation
that arises from children born after migration. Consequently, it is more likely that the
correlation between graduation in the S-system and migration reflects a non-endogenous
causality.
Table 8 shows the results of the reduced-form and structural probability models (first and
third steps).41 The reduced form model indicates that migration is more likely amongst the
41
Selectivity corrected wage equations and the distribution of the potential wage differential are available by
the authors upon request.
33
older individuals within this age bracket, and less likely amongst women. As expected,
children born before the migration time span discourage migration, while the ethnic
background does not play any role in explaining recent migration. The log of the state
population at origin seems to affect the migration probability in the expected direction,
although we cannot be sure that the statistical insignificance of this coefficient is due to the
multicolinearity associated with the level of aggregation of the variable relative to the
number of positive outcomes.
Of more interest are the variables that can influence the migration decision through their
impact on labour earnings differentials. The reduced form shows that regarding education,
migrants are selected from the upper and lower tail of the distribution. Although statistically
insignificant, it is interesting that this selectivity pattern tends to vanish after controlling for
the impact that education causes through the earnings differential. Also curious is the fact
that the impact of being a S-system graduate does not disappear after controlling for its
contribution to the earnings gap (although its impact slightly declines). This means that the
impact of the S-system on migration is not driven by the earnings gap caused by the
differential returns that the training has in states of origin and destination, but by its
contribution to enhancing non-observables such as information, cultural endowments
(networks) and the cumulative causation amongst S-system graduates (S-system graduates
share specific information that encourages additional migration flows) that strongly
determines migration (see Tilly and Brown, 1967; Lomnitz, 1977; Massey, 1990). The
structural model confirms the fact that the number of children born before the migration
time span contributes to increasing the cost term of the migration utility function (Sjaastad
1962, Becker 1962).
The conclusion is that S-system graduates are, on average, more likely to migrate than their
non-trained counterparts. At the country level and based on the same highly restricted nonenrolled employed population (those aged between 30 and 40, without children aged less
than 5), the S-system explains, holding everything else constant, the additional migration
flow of 82,000 individuals between states which is significant at the 10 percent level. If we
allow for unemployment, the “S-system” migration flow rises up to 106,000 individuals.
Generalizations arguing that this type of education unambiguously discourages social
mobility should therefore be carefully formulated.
34
Table 8: Probit model: determinants of inter-state Migration (Reduced form and structural)
Variables / Model
Earnings differential
Earnings differential – squared
S-system
Age
Incomplete basic education
Complete basic education
Incomplete secondary education
Complete secondary education
Incomplete secondary education
Complete secondary education
Average household education
Female
Children older than 5
Ethnicity 1 (white)
Ethnicity 2 (mulatto)
Unemployment rates at origin (state level)
Log of the state population at origin
Constant
Federal Unit controls at origin
Observations
Population size
F
Prob > F
Dependent variable: recent inter-state migration
(Five years time span)
Aged 30-40
Reduced
Structural
0.127*
(0.075)
-0.010*
(0.006)
0.023
(0.088)
-0.062
(0.099)
-0.119
(0.124)
-0.103
(0.096)
-0.020
(0.144)
0.088
(0.101)
0.006
(0.009)
-0.113***
(0.037)
-0.135***
(0.021)
0.022
(0.070)
0.067
(0.068)
-0.033*
(0.020)
-0.090
(0.056)
0.526
(1.211)
Yes
0.074
(0.109)
-0.136
(0.155)
0.123*
(0.074)
-0.010
(0.006)
0.019
(0.088)
-0.068
(0.100)
-0.136
(0.127)
-0.106
(0.096)
-0.020
(0.144)
0.064
(0.111)
0.008
(0.009)
-0.092**
(0.045)
-0.136***
(0.021)
0.033
(0.074)
0.080
(0.069)
0.037***
(0.014)
-0.034
(0.091)
-2.092
(1.396)
Yes
29871
13962307
5.164
0.000
29871
13962307
4.064
0.0000
Excluded categories: no education, agriculture, managing directors and CEOs, men, other ethnic backgrounds.
Source: Own calculations based on PNAD 2007.
35
6. Shifts in the demand for labour and income inequality
So far we have demonstrated the SENAI system can offer substantial labour market returns
to participants. At the same time, it is useful to relate these findings to more general trends
in the Brazilian labour market. In particular, can SENAI contribute to explaining the fact that
the Brazilian economy was able to reduce general and youth unemployment from 12.4 and
25.3 percent in 2003 to 5.4 and 13.7 percent in 2012 respectively?42 What is the role of the
SENAI in this, in the words of Amann (2011), second Brazilian miracle? And what is the
impact of the SENAI training on changes in income inequality, particularly the substantial
decline in income inequality since the mid-1990s in Brazil? Our empirical analysis suggests
that the SENAI has contributed to this story of success by providing valuable skills to the
labour force. However, the links to inequality reduction are not straightforward.43 Firstly, the
contribution of the SENAI in shaping the skill’s endowment is somewhat ambiguous. As of
2007, one fifth of the S-system trained population had less than a basic education, while
about 50% achieved at least a secondary education (see Table 2 above). Secondly, given the
idea that the expansion of education is a continuous, cumulative and highly persistent
process, we argue that labour and educational supply factors can hardly explain pronounced
inequality changes over relatively short periods of time. Instead, there is evidence that shifts
in the demand for labour plays a prominent role in explaining inequality dynamics in Brazil.
For instance, starting in the 1970s, a shift in the labour demand away from low-skilled
occupations caused relative wages for less-educated individuals to decline and labour
income inequality to rise, as they were unable to move towards the capital intensive growing
sectors (see Cimoli and Katz, 2002). More recently, the change in the labour demand
favouring less-skilled individuals may explain the apparent paradox of low unemployment
rates and low, however improving, achievements in education as well as the sharp decline in
labour income inequality (see Barros et al. 2010; Schwartzman and Christophe, 2005;
Schwartzman and de Moura Castro, 2013).44
Recent literature suggests that demand shifts may account for a substantial part of the
observed distributional changes in the household income distribution. For instance, Klasen
et al. (2014) offer a plausible explanation of the inequality dynamics in Honduras over the
period 1990-2007. Thus, the same transmission mechanisms appear to be plausible
candidates to explain the linkages between demand shift and income inequality. Although it
is unlikely that the SENAI could have played a central role in determining inequality trends, it
is likely that the SENAI improves labour mobility and thus, reduces the potential
disequalizing shifts in labour demand. Klasen et al. (2014) and Villalobos Barría (2012)
argued that shifts away from the traditional (less-skilled) sector together with an immobile
42
See Banco Central Do Brasil (2013) and ILO (2013) for total and youth unemployment figures respectively.
According to Barros et al. (2010) the fall in labour income inequality explains about half of the decline in
overall income inequality (household per capita income) during the 2001-2007 period. The authors argue that
an accelerated expansion of education and declining returns to education largely account for the observed
declining household income inequality trend.
44
Schwartzman and de Moura Castro (2013) argue that job creation mainly responds to an increasing demand
for low-skilled workers in the service sector. Schwartzman and Christophe (2005) argue that new technologies
tend to favour two groups: One group providing standardized services that requires a low level of skills, and
another group supplying non-standardized services, such as sale, and other activities that demand a higher
level of skills.
36
43
working force might result in a strong disequalization of the labour incomes while increasing
informality (see also Soares et al., 2003).
7. Conclusion
With the challenges of skills shortages, substantial youth unemployment, and a demographic
bulge in Sub-Saharan African countries in mind, the main purpose of this paper has been to
understand the determinants of participation in Brazil’s S-system (mainly SENAI) as well as a
thorough examination of the measurable consequences of S-system participation on
facilitating the school-to-work transition.
This study has provided an overview of potential and limitations of the S-system of
vocational training system in Brazil. Our impact evaluation suggests that the weaknesses of
the S-system originate from the same source that makes it effective. The search for
matching supply and demand and its decentralised framework determines that the
institution can be as good as the local employers’ associations are.
Our analysis, based in five different models, sheds some light on the dimensions that need to
be taken into consideration when evaluating such a complex and decentralised institution.
These dimensions are all important and worth considering (without a hierarchic order) in
Africa independently of the country-specific context.
In the racial dimension, policymakers have to ensure that any intervention to create or
modify the vocational training system needs to be oriented around offering each ethnic
group the same development chances. In the financing dimension, the focus should be on
finding the proper structure that avoids fluctuations and uncertainties. In the productivity
dimension, the emphasis has to be on improving labour market outcomes. In the social
mobility dimension, the set of interventions has to be oriented around reducing the
segmentation in occupations, in particular towards low-skilled activities and in promoting
new chances of internal migration. In the gender dimension, efforts have to be devoted to
fighting discrimination promoting disadvantaged groups. Finally, in the geographic
dimension, the system has to cover the whole country, promoting a reduction in the speed
of urbanization and the formation of overpopulated cities surrounded by violence and low
standards of living.
The sustainability of the S-system and of the SENAI lies on its good performance in most
dimensions. In this study we did not find any signal of ethnic inequality caused by the Ssystem. The financing schema proved to balance a market driven component with a stable
public funding. Regarding the geographic dimension, the S-system is present in the 27 states
of the country, and the region of residence does almost not affect the enrolment probability.
The gender and social mobility dimensions are the spaces where the S-system and the SENAI
in particular could have done more over the past decades. It is not clear that the male bias of
the S-system is exclusively a demand driven issue. In the same way, although the S-system
encourages geographical mobility, this is not sufficient to meeting the needs of the poor or
to encourage the social mobility of the disadvantaged. It will probably be argued that such
37
challenges are not specific to the S-system. The worldwide leadership of the system imposes
the responsibility to innovate in both dimensions as the system represents the model to
follow in Africa.
38
References
Amann, E. (2011). A new Brazilian economic miracle?: Brazil (Focus). CESifo forum : a
quarterly journal on European economic issues.
Arriagada A. M., and A. Ziderman (1992). Vocational Secondary Schooling, Occupational
Choice, and Earnings in Brazil. Policy Research Working Paper. Education and
Emploment Division, Population and Human Resoruces Department. The World
Bank.
Assaad, R., and F. Roudi-Fahimi (2007). Youth in the Middle East and North Africa:
Demographic opportunity or challenge? Reference Bureau MENA Policy Brief,
April.
Atchoarena, D., and A. Delluc (2001). Revisiting technical and vocational education in subSaharan Africa: Provision, patterns and policy issues. UNESCO-IIEP, Paris.
Audas, R., É. Berde, and P. Dolton (2005). “Youth unemployment and labour market
transitions in Hungary”,Education Economics, 13, pp. 1-25.
Baer, W. (2007). The Brazilian Economy: Growth and Development, 7th edition, Boulder:
Lynne Rienner.
Banco Central do Brasil (2013). Boletim Regional do Banco Central do Brasil.
Barros, R., M. De Carvalho, S. Franco, and R. Mandonca (2010). “Markets, the State, and the
Dynamics of Inequality in Brazil”. In L.F. López-Calva and N. Lustig (eds), Declining
Inequality in Latin America: A Decade of Progress? New York and Washington,
DC: UNPD and Brookings Institution Press, 134-74.
Becker, G. S. (1962). Investment in Human Capital: A Theoretical Analysis. J. of Political
Economy, Supplement 70: 9-49.
Biavaschi, C., W. Eichhorst, C. Giulietti, M. J. Kendzia, A. Mravyev, J. Pieters, N. RodriguezPlanas, R. Schmidl, and K. F. Zimmermann (2012). Youth Unemployment and
Vocational Training, IZA Discussion Paper No. 6890, IZA Bonn.
Borjas, G. (1987). Self-Selection and the Earnings of Immigrants. American Economic Review
77: 531-53.
Borjas, G. (1988). Immigration and Self-Selection. NBER Working Paper No. w2566,
http://www.nber.org/papers/w2566. Accessed 10 July 2011.
Bratberg, E., and O. Nilsen (1998). Transition from school to work: Search time and job
duration, Institute for the Study of Labor (University of Bonn) Discussion Paper
27.
Breen, R., and J. Jonsson (2007). “Explaining change in social fluidity: educational
equalization and educational expansion in twentieth-century Sweden”, American
Journal of Sociology 112:1775–810.
Brücker, H., and P. Trübswetter (2007). Do the Best Go West? An Analysis of the SelfSelection of Employed East-West Migrants in Germany. Empirica 34: 371-395.
39
Buchmann, C., H. Park (2009). “Stratiﬁcation and the Formation of Expectations in Highly
Differentiated Educational Systems.” Research in Social Stratiﬁcation and
Mobility 27:245–67.
Cardoso, A. (2008). Transições da escola para o trabalho no Brasil: persistência da
desigualdade e frustração de expectativas. Dados, v.51, n.3, p.569-616, 2008.
CASEN (Encuesta Nacional de Caracterización Socio-económica) (n.d) Chilean Household
surveys.
CEDEFOP (1998). Who pays for training? Some policy approaches to financing vocational
training. European Vocational Training Journal, No. 13. European Centre for the
Development of Vocational Training.
Chigunta, F., N. Chisup, and S. Elder (2013). Labour Market Transitions of Young Women and
Men in Zambia. Work4Youth publication series No.4, International Labour Office,
Geneva: ILO
Cimoli, M., and J. Katz (2002). “Structural Reforms, Technological Gaps and Economic
Development: A Latin American Perspective”, Serie Desarrollo Productivo n. 129,
ECLAC, Santiago Chile.
CINTEFOR/OIT (1990). La formación profesional en el umbral del siglo XXI. M.A. Ducci (ed.)
CINTERFOR, Montevideo, Uruguay.
CNI (2007). Educação para a nova indústria: uma ação para o desenvolvimento sustentável
do Brasil. SENAI, Confederação Nacional da Indústria. Brasilia
Colistete, R. P. (2010). “Revisiting import-substituting industrialisation in post-war
Brazil,MPRA Paper N° 24665. en: http://mpra.ub.unimuenchen.de/24665/posted 30. August 2010(consulta realizada 17/07/2013).
De Mel, S., S. Elder, and M. Vansteenkiste (2013). Labour Market Transitions of Young
Women and Men in Liberia. Work4Youth publication series No.4, International
Labour Office, Geneva: ILO
Dehejia, R. H. (2005). “Practical Propensity Score Matching: A Reply to Smith and Todd,”
Journal of Econometrics 125 (March–April 2005), 355–364.
DFID (2007). Technical and vocational skills development, Briefing. Department for
International Development. United Kingdom.
De Moura Castro, C. (1979). Vocational Education and the Training of Industrial Labour in
Brazil, Intemational Labour Review, Vol. 118, No 5, September-October.
De Moura Castro, C., and A. Verdisco. (1998). “Training Unemployed Youth in Latin America:
Same old sad story?”, Inter-American Development Bank, Washington, DC.
De Moura Castro, C. (2011). Learning and Occupation, Practices and Policies. Klaus Schwarz
Verlag. Berlin
ECLAC (2004). Desarrollo productivo en economías abiertas. ECLAC, Executive Secretary, June
2004.
40
Ferreira, P., and Facchini, G. (2005). “Trade liberalization and industrial concentration:
Evidence from Brazil.” The Quarterly Review of Economics and Finance, 45, (23),
pp. 432 – 446.
Forte Aguas, M. (2011). Ensino Profissional e Rendimento do Trabalho: Uma Aálise para o
Brasil. Mercado de Trabalho. 47. pp 17-28.
Fresneda, B. (2009). Transição da escola para o trabalho e estratificação social. Segurança
Urbana e Juventude, Araraquara, v.2, n.1/2, 2009.
Fresneda, B. (2012). Desigualdade de Oportunidade no Ensino Médio Técnico Brasileiro.
Mimeo. Presented at “36º Encontro Anual da Anpocs”. Águas de Lindóia, Sao
Paulo.
Gasskov, V. (1994). Revenue-generating levy schemes. In Gasskov, V. (ed.) Alternative
schemes of financing training. International Labour Office, Geneva.
García, M., and J. Fares (2008a). Transition to Working Life for Africa’s Youth. In: Garcia,
Marito; Fares, Jean (eds.): Youth in Africa’s Labor Market, World Bank, 15-25.
García, M., and J. Fares (2008b). Why is it important for Africa to invest in its youth? In:
Garcia, Marito; Fares, Jean (eds.): Youth in Africa’s Labor Market, World Bank, 314.
Giulietti C., G. Ning, and K.F. Zimmermann (2012). Self-employment of rural-to-urban
migrants in China. International Journal of Manpower, 33: 96-117.
Goldthorpe, J. H., and R. Erikson (1992). The Constant Flux: A Study of Class Mobility in
Industrial Societies. Oxford: Oxford University Press.
Goldfeld, S., and R. Quandt (1973). The Estimation of Structural Shifts by Switching
Regressions. Annals of Economic and Social Measurement 2: 475-85.
Greenwood, M. J. (1993). Migration: A Review. Regional Studies 27: 295-296.
Grubb, W.N. (1985). The Convergence of Educational Systems and the Role of Vocationalism,
Comparative Education Review, 29, No. 4, pp. 526-548.
Hasenbalg, C. (2003). A transição da escola ao mercado de trabalho. In: HASENBALG, C.;
SILVA, N. V. (Org.). Origens e destinos: desigualdades sociais ao longo da vida. Rio
de Janeiro: Iuperj: TopBooks, 2003. p.147-172.
Hirano, K., G. W. Imbens, and G. Ridder (2003). Efficient estimation of average treatment
effects using the estimated propensity score. Econometrica 71: 1161–1189.
Horvitz, D. G., and D. J. Thompson (1952). A generalization of sampling without replacement
from a finite universe. Journal of the American Statistical Association 47: 663–
685.
Hove, M. (2013). The Urban Crisis in Sub-Saharan Africa: A Threat to Human Security and
Sustainable Development. Stability, 2(1): 7, pp. 1-14, DOI:
http://dx.doi.org/10.5334/sta.ap
41
ILO (2013). Global Employment Trends for Youth 2013, A generation at risk. International
Labour Office. Geneva.
Instituto de Desenvolvimento da Guanabara (1970). Avaliacao do SENAI-GB. Rio de Janeiro.
Klasen, S., Otter, T., and Villalobos Barría, C. (2014). The Dynamics of Inequality Change in a
Highly Dualistic Economy. In: G.A. Cornia (ed.), Falling Inequality in Latin
America. Policy Changes and Lessons. UNU-WIDER Studies in Development
Economics. Oxford University Press.
Kniivilä, M. (2007). Industrial development and economic growth: Implications for poverty
reduction and income inequality. In: Industrial Development for the 21st Century:
Sustainable Development Perspectives. United Nations, Department of Social and
Economic Affairs. New York, USA. Pp. 295-332.
Kingstone, P. (2012). The Brazilian Miracle and Its Limits. Law and business review of the
Americas.
Koenker, R., and G. Bassett (1978). Regression Quantiles. Econometrica, Vol. 46(1), pp. 3350.
Lassibille, G., L. N. Gomez, I. A. Ramos, and C. D. Sanchez. (2001). Youth transition from
school to work in Spain, Economics of Education Review, 20(2), pp. 139-149.
LDES (n.d). Labour Demand Enterprise Survey. Work4Youth Project. International Labour
Organization.
Lomnitz, L. (1977). Networks and Marginality: Life in a Mexican Shantytown. Academic Press,
New York.
Maddala, G. (1983). Limited-dependent and Qualitative Variables in Econometrics.
Cambridge University Press. Cambridge.
Maddala, G., and F. Nelson (1975). Switching Regression Models with Exogenous and
Endogenous Switching. Proceedings of the American Statistical Association,
Business and Economic Statistics Section, 423-426.
Maddison, A. (1995). Monitoring the World Economy, 1820-1992. Paris: OCDE.
Massey, D. S. (1990). Social Structure, Household Strategies, and the Cumulative Causation
of Migration. Population Index 56: 3-26.
McVicar, D., and M. Anyadike-Danes (2000). Predicting successful and unsuccessful
transitions from school to work using sequence methods. Northern Ireland:
Northern Ireland Economic Research Centre, 2000. (Working Paper Series, n.55).
Mizen P. (2004). The Changing State of Youth. Basingstoke, Palgrave, pp. 209 + xviii. ISBN
0333739507.
Müller, W., and M. Gangl (2003). The Transition from School to Work: A European
Perspective, in: W. Müller and M. Gangl (eds), Transition from Education to Work
in Europe. The Integration of Youth into EU Labour Markets, Oxford University
Press, Oxford, pp. 1–22.
42
Müller, W., M. Gangl, and D. Raffe (2003). Conclusions: explaining cross-national differences
in school-to-work transitions. In: Müller, W., Gangl, M (Org.). Transitions from
education to work in Europe. Oxford: Oxford Scholarship Online, 2003. p.277335.
Müller, W., and R. Pollak (2005). Social Mobility in Western Germany: The Long Arms of
History Discovered? In: BREEN, R. (Ed.). Social Mobility in Europe. Oxford: Oxford
University Press.
Mussa, R. (2013). Labour Market Transitions of Young Women and Men in Malawi.
Work4Youth publication series No.4, International Labour Office, Geneva: ILO.
OECD, CPRN (2005). From Education to Work: A Difficult Transition For Young Adults With
Low Levels of Education. Paris.
Oketch, M. (2007). To vocationalise or not to vocationalise? Perspectives on current trends
and issues in technical and vocational education and training (TVET) in Africa”,
International Journal of Educational Development, 27, 220-234.
PNAD (Pesquisa Nacional por Amostra de Domicílios) (n.d). Brazilian Households Survey.
Instituto Brasileiro de Geografia e Estatística (IBGE).
Recieta Federal (2008). Contribuições Devidas a Terceiros Coletânea da Legislação. COSIT,
Cotri, Ditri.
Ribeiro, C. A. C. (2007). Estrutura de classe e mobilidade social no Brasil. Bauru: EDUSC.
Robins, J. M., and A. Rotnitzky (1995). Semiparametric efficiency in multivariate regression
models with missing data. Journal of the American Statistical Association 90:
122–129.
Robins, J. M., A. Rotnitzky, and L. P. Zhao (1994). Estimation of regression coefficients when
some regressors are not always observed. Journal of the American Statistical
Association 89: 846–866.
Robins, J. M., A. Rotnitzky, and L. P. Zhao (1995). Analysis of semiparametric regression
models for repeated outcomes in the presence of missing data. Journal of the
American Statistical Association 90: 106–121.
Rodrigues, J. (1998). O moderno príncipe industrial: o pensamento pedagógico da
Confederação Nacional da Indústria. Campinas: Editora Autores Associados.
Rogers, W. H. (1992). Quantile regression standard errors. Stata Technical Bulletin 9, pp. 1619.
Roy A.D. (1951). Some Thoughts on the Distribution of Earnings. Oxford Economic Papers 3:
135-146
Sjaastad, L. (1962). The Costs and Returns of Human Migration. The Journal of Political
Economy 70: 80-93.
Schwartzman S., Bomeny H. M. B. and V.M.R. Costa (2000). Tempos de Capanema. 2ed São
Paulo. Paz e Terra; Editora FGV, 2000.
43
Schwartzman S., and M. Christophe (2005). A sociedade do conhecimento e a educacao
tecnológica. SENAI/DN, Brasilia.
Schwartzman S., C. de Moura Castro (2013). Ensino, formação professional e a questão da
mão de obra. aval. pol. públ. Educ., Rio de Janeiro, v. 21, n. 80, p. 563-624.
SENAI-DR/RJ (1979). Avaliação da metodologia da Formacao individual adotada pelo DR/RJ
SENAI-DN (2000). Plano estratégico do SENAI; Revisão 2000-2010. Brasilia.
SENAI-DN (2011). Departamento Nacional. Programa de Acompanhamento de Egressos do
SENAI. Avaliação Triênio 2008/2010. Brasília.
SENAI-DN (2012). SENAI actions for sustainable development. Brasilia.
Sistema Firjan (2013). Orçamento e Execuçao Orçamentária 2013 – SENAI/DR-RJ. Lei de
Diretrizes Orçamentárias,
Shavit, Y., and W. Müller (1998). From School to Work: A Comparative Study of Educational
Qualifications and Occupational Destinations. Clarendon, Oxford.
Soares, S., L. Carvalho, and B. Kipnis (2003). Os jovens adultos de 18 a 25 anos: retrato de
uma dívida da política educacional. Texto de Discussão, n.954. Rio de Janeiro:
IPEA.
STWS (n.d). School-to-Work Transition Survey. Work4Youth Project. International Labour
Organization.
Tilly, C., and C.H. Brown (1967). On Uprooting, Kinship, and Auspices of Migration.
International Journal of Comparative Sociology 8: 139-164.
UNPF (2007). State of World Population 2007: Unleashing the Potential of Urban Growth.
United Nations Population Fund.
Vasconcellos, L., F. C. Lima, and N. Menezes-Filho (2010). Avaliação Econômica do Ensino
Médio Profissional. Relatórios de Avaliação Econômica n. 14, Fundacao Itaú
social.
Vera, A. (2009). Los jovenes y la formación para el trabajo en América Latina. Documento de
Trabajo N° 25. Centro de Implementación de Políticas Públicas para la Equidad y
el Crecimiento.
Vikström J. (2014). IPW estimation and related estimators for evaluation of active Labor
market policies in a dynamic setting. Working Paper 1. Institute for Evaluation of
Labour Market and Education Policy.
Villalobos Barría, C. (2012). Sectorial Shifts and Inequality. How to Relate Macroeconomic
Events to Inequality Changes. IAI Discussion Papers 219, Ibero-America Institute
for Economic Research.
Wilson, D. N. (1993). The effectiveness of national training boards. Geneva: International Labour
Organization, Training Discussion Paper No. 110.
Wilson, D. N. (2006). Development of the SENAI Post-Secondary Sector in Brazil. Higher
Education Perspectives. Special Issue, March 2006, p. 82-95.
44
Wooldridge, J. M. (2002). Inverse probability weighted M-estimators for sample selection,
attrition, and stratification. Portuguese Economic Journal 1: 117–139.
Wooldridge, J. M. (2007). Inverse probability weighted estimation for general missing data
problems. Journal of Econometrics 141:1281–1301.
Wooldridge, J. M. (2010). Econometric Analysis of Cross Section and Panel Data. 2nd ed.
Cambridge, MA: MIT Press.
World Bank (1989). Sub-Saharan Africa: From Crisis to Sustainable Growth. Washington DC:
World Bank.
Zhao, Z. (2004). “Using Matching to Estimate Treatment Effects: Data Requirements,
Matching Metrics, and Monte Carlo Evidence,” Review of Economic and Statistics
86 (February 2004), 91–107.
45
APPENDIX
Table A.1: Probability model for the enrolment in a technical or technological training course,
2007
Population
Age
Incomplete basic
Complete basic
Complete secondary
Incomplete tertiary
Complete tertiary
Female
Ethnic I (white)
Ethnic II (mulatto)
Spouse
Children
Household size
Northeast
Southeast
South
Midwest
Employed
Formal
Union membership
Hours of work
Manufacturing
Other Services
Managers
Professionals
Technicians
Clerks
Service workers
Craft and trade workers
Operators
Constant
Observations
Population
Design df
F-statistic
p-value
Number of Strata
Number of PSU
Dependent variable: enrolment in a technical and technological training course, by type of institution
Sample 1
Sample 2
Sample 3
Other
Other
Other
S-system
S-system
S-system
institutions
institutions
institutions
-0.007***
-0.013***
-0.029***
-0.032***
-0.030***
-0.032***
0.260**
0.369***
0.200
0.368***
0.141
0.416***
0.747***
0.746***
0.501**
0.717***
0.452**
0.730***
0.911***
0.921***
0.777***
0.903***
0.716***
0.891***
0.940***
1.036***
0.861***
1.102***
0.799***
1.055***
0.490***
0.569***
0.353*
0.594***
0.279
0.594***
0.611***
0.802***
0.331
0.873***
0.272
0.837***
-0.320***
-0.110**
-0.397***
-0.196***
-0.399***
-0.198***
0.016**
0.012***
0.022**
0.006
0.008
-0.002
-0.203***
0.073***
-0.308***
0.029
-0.324***
0.020
-0.035
-0.037
-0.066
-0.036
-0.052
-0.043
-0.003
-0.024
0.010
0.003
0.018
0.010
0.096**
0.024
0.132***
0.045*
0.143***
0.061**
-0.085
-0.051*
-0.140
-0.050
-0.115
-0.055
0.050
0.077***
-0.028
0.054
0.005
0.066
-0.028
0.021
-0.120
-0.011
-0.088
-0.017
-0.035***
-0.037***
-0.048***
-0.034***
-0.044***
-0.030***
0.058
-0.099***
0.083
-0.115**
0.020
-0.112**
-0.050
0.053
0.005
0.047
0.012
0.058
0.041
0.207***
0.122
0.168***
0.103
0.166***
0.163
0.232***
0.184*
0.175***
0.182
0.170***
0.055
0.075
0.106
0.048
0.038
0.016
0.060
0.288***
0.043
0.245***
0.020
0.215***
0.137
-0.293***
0.111
-0.477***
0.071
-0.575***
0.133***
0.037
-0.002***
0.033
0.107**
-0.025
0.061
0.172**
0.078
0.027
0.168
0.120
0.196**
0.243**
0.275***
-2.985***
174872
83282219
5455
18.276
0.000
637
6092
-0.081***
0.117***
-0.001***
0.309***
0.129***
0.078*
0.154***
0.150***
0.387***
0.441***
0.538***
0.423***
0.459***
0.433***
0.318***
-2.318***
180305
85990337
5493
82.743
0.000
699
6192
0.224***
0.072
-0.003***
-0.003
0.068
-0.105
-0.070
0.008
0.180
0.248
0.431**
0.289*
0.339**
0.484***
0.360**
-2.316***
67046
31161384
4999
10.278
0.000
627
5626
-0.073**
0.147***
-0.001***
0.555***
0.114***
0.039
0.130***
0.117*
0.403***
0.601***
0.652***
0.593***
0.610***
0.570***
0.416***
-1.810***
70351
32768719
5040
40.617
0.000
649
5689
0.245***
0.071
-0.004***
-0.190
0.064
-0.096
-0.043
0.018
0.264
0.323
0.514*
0.369
0.413
0.549**
0.425
-2.137***
54798
25145643
4658
8.717
0.000
592
5250
-0.066*
0.159***
-0.001***
0.566***
0.111***
0.012
0.115**
0.131**
0.495***
0.688***
0.730***
0.671***
0.660***
0.630***
0.466***
-1.738***
57747
26567877
4691
27.284
0.000
592
5283
Excluded Categories: Male, other ethnic groups, heads of household, and northern region, non-enrolled, unemployed.
Note: Samples (1) consist of individuals aged + 10. Sample (2) is (1) aged between 15 and 29. Sample (3) is sample (2) in urban areas.
Sample (4) corresponds to women of (3).
46
Table A.2: Probability model for the enrolment in a professional qualification training course,
2007
Dependent variable: enrolment in a professional qualification training course, by type of institution
Population
Sample 1
Sample 2
Sample 3
S-system
Other
institutions
Age
-0.014***
-0.014***
-0.039***
-0.023***
-0.037***
Years of schooling
0.094***
0.116***
0.085***
0.082***
0.081***
0.072***
-0.466**
-0.245***
-0.743**
-0.243***
-0.716**
-0.225***
0.028***
S-system
Other
institutions
S-system
Other
institutions
-0.023***
0.019*
0.016***
0.011
0.031***
-0.001
-0.297***
0.009
-0.366***
0.007
-0.339***
-0.000
Ethnic I (white)
-0.004
-0.108***
-0.006
-0.101**
-0.011
-0.095**
Ethnic II (mulatto)
-0.088
-0.068*
-0.038
-0.082*
-0.045
-0.076
0.093
-0.012
0.101
0.007
0.158**
0.017
Spouse
0.003
-0.076**
0.178
-0.037
0.163
-0.053
Children
0.260***
0.106***
0.321***
0.171***
0.353***
0.175***
0.167
-0.046
0.142
0.002
0.198
0.005
-0.071***
-0.026***
-0.087***
-0.042***
-0.095***
-0.047***
0.186
-0.017
0.195
-0.077
0.204
-0.048
Northeast
-0.182
-0.019
-0.112
0.016
0.037
0.058
Southeast
0.068
0.127***
0.201
0.195***
0.306**
0.216***
South
0.069
0.188***
0.174
0.251***
0.271**
0.258***
Midwest
0.054
-0.114**
0.152
-0.147**
0.210
-0.124*
-0.084
0.147***
-0.082
0.062*
-0.075
0.059*
Employed
-0.023
-0.345***
0.132
-0.605***
-0.052
-0.580*
0.043
Female
Household size
Formal
0.137**
-0.011
0.255***
0.033
0.284***
Union membership
-0.043
0.018
0.064
0.001
0.048
0.021
Hours of work
-0.001
-0.000
-0.003***
-0.000
-0.003***
-0.000
0.454**
0.473***
0.000
0.604***
0.000
0.577***
Manufacturing
0.259***
0.184***
0.334***
0.185***
0.295***
0.160***
-0.049
0.099*
-0.279
0.102
-0.338
0.069
0.177
0.163***
0.243
0.200***
-
0.164**
Other Services
0.104
0.038
0.105
0.053
-
0.041
Managers
-0.142
0.162
-0.147
0.364
-
0.343
Professionals
-0.443*
-0.047
-0.348
0.242
-
0.242
Technicians
0.078
0.496***
-0.001
0.669***
0.314
0.622*
Clerks
-0.001
0.391***
-0.011
0.624***
0.235
0.614*
Service workers
0.123
0.419***
0.096
0.560***
0.349*
0.526*
Craft and trade workers
0.073
0.307**
0.205
0.473**
0.419**
0.397
Operators
0.223
0.284**
0.199
0.463**
0.438**
0.425
-3.565***
-3.297***
-2.940***
-2.714***
-2.987***
-2.578***
Constant
Observations
Population
320884
322613
100313
101868
82985
84368
151363832
152216286
46216520
46971071
37747149
38412456
Design df
6001
6032
5415
5430
5017
5025
F-statistic
35.202
130.673
16.292
50.004
12.729
34.919
p-value
0.000
0.000
0.000
0.000
0.000
0.000
Number of Strata
682
753
667
695
629
628
Number of PSU
6683
6785
6082
6125
5646
5653
Excluded Categories: Male, other ethnic groups, heads of household, and northern region, non-enrolled, unemployed.
47
Table A.3: IPW average treatment effects of vocational training on labour monthly earnings
by institution, household position and aggregation samples (1-4), Reais of 2007
and percentages over non-trained individuals.
Professional
qualification
Absolute
effect
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Technical and
Technological
levels
Whole
Heads
Other
S-system
institutions
68
117
157
52
84
40
-17
16
Whole
Other
S-system
institutions
8.0
13.6
26.5
8.7
13.2
6.3
-2.9
2.6
Other
S-system
institutions
99
179
257
50
120
47
Heads
Other
S-system
institutions
9.8
17.5
38.2
7.3
16.3
6.4
-
Whole
Heads
Other
institutions
32
67
23
60
0
37
Children
Other
S-system
institutions
4.9
10.2
4.0
10.3
0.0
5.9
S-system
Children
Absolute
effect
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
65
235
104
106
86
105
106
18
148
303
40
-
71
160
175
-
158
126
143
-
116
83
77
-
Whole
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
S-system
7.4
38.8
15.9
17.8
Other
institutions
8.9
16.1
15.4
2.8
Whole
All levels
Children
Absolute
effect
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Other
institutions
68
109
168
63
87
54
0
16
Whole
Other
S-system
institutions
7.9
12.2
28.3
10.4
13.7
8.4
0.0
2.7
S-system
Heads
S-system
13.9
44.0
5.3
-
Other
institutions
6.0
20.9
21.1
-
Heads
Other
institutions
104
149
262
72
110
73
Heads
Other
S-system
institutions
10.3
13.9
38.9
10.3
14.9
9.7
S-system
Children
S-system
23.7
21.5
22.7
-
Other
institutions
16.4
13.3
11.7
-
Children
Other
institutions
52
79
40
65
24
46
Children
Other
S-system
institutions
7.9
11.8
6.9
11.0
3.8
7.3
S-system
Note: Samples (1) consist of individuals aged + 10. Sample (2) is (1) aged between 15 and 29. Sample (3) is sample (2) in urban
areas. Sample (4) corresponds to women of (3).
48
Table A.4: IPW average treatment effects of vocational training on monthly working hourks
by institution, household position and aggregation samples (1-4), hours and
percentages over non-trained individuals.
Professional
qualification
Absolute
effect
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Technical and
Technological
levels
Absolute
effect
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Whole
Heads
Other
S-system
institutions
0.6
2.4
-1.1
2.9
-1.3
2.5
-2.3
2.7
Whole
Other
S-system
institutions
0.3
1.3
-0.6
1.6
-0.7
1.4
-1.4
1.6
Other
S-system
institutions
2.6
4.4
-3.6
2.9
-2.7
0.9
Heads
Other
S-system
institutions
1.4
2.3
-1.9
1.5
-1.4
0.5
-
Whole
Heads
Other
institutions
-1.3
-0.3
-0.8
-1.2
4.6
-1.5
12.7
-0.3
Whole
Other
S-system
institutions
-0.7
-0.2
-0.4
-0.7
2.5
-0.8
7.5
-0.2
S-system
All levels
Sample 1
Sample 2
Sample 3
Sample 4
Other
institutions
0.9
3.5
2.7
3.5
-2.8
2.8
Children
Other
S-system
institutions
0.5
1.9
1.5
1.9
-1.5
1.5
S-system
Children
Other
institutions
2.6
1.7
-5.3
0.0
-5.4
-2.0
Heads
Other
S-system
institutions
1.4
0.9
-2.7
0.0
-2.8
-1.1
S-system
Whole
Absolute
effect
Children
Other
institutions
3.1
-2.4
-0.9
-0.4
4.7
0.1
Children
Other
S-system
institutions
1.7
-1.3
-0.5
-0.2
2.6
0.0
S-system
Heads
S-system
Other
institutions
0.4
-1.1
-0.4
-0.2
1.7
2.0
1.6
2.0
Children
S-system
Other
institutions
S-system
Other
institutions
2.6
-3.8
-3.0
-
3.6
2.3
0.3
-
1.2
2.1
-1.5
-
2.0
2.6
2.2
-
Whole
Heads
Children
Relative effect
(%)
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
0.2
-0.6
-0.2
-0.1
0.9
1.1
0.9
1.2
1.4
-2.0
-1.6
-
1.9
1.2
0.2
-
0.7
1.2
-0.8
-
1.1
1.4
1.2
-
49
Table A.5: IPW average treatment effects of vocational training on hourly wages by
institution, household position and aggregation samples (1-4), Reais of 2007
and percentages over non-trained individuals.
Whole
Professional
qualification
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Heads
S-system
Other
institutions
0.34
0.84
0.28
-0.22
0.59
0.24
0.11
-0.01
S-system
S-system
Other
institutions
0.46
1.59
0.69
-
0.73
0.20
0.22
-
0.45
-0.22
-0.34
-
0.25
0.21
-0.02
-
Whole
Heads
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
6.3
22.4
7.1
-5.2
10.6
6.3
2.7
-0.3
7.4
41.4
16.4
-
11.4
5.1
5.3
-
10.6
-5.8
-8.4
-
5.7
5.4
-0.5
-
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Heads
S-system
Other
institutions
0.02
1.01
0.14
-0.34
0.64
0.77
0.75
-0.05
Children
S-system
Other
institutions
S-system
Other
institutions
0.47
1.68
0.18
-
0.17
1.67
1.97
-
1.09
0.54
0.38
-
1.30
0.54
0.48
-
Whole
Heads
Children
Relative effect
(%)
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
0.3
26.6
3.4
-7.8
10.2
18.5
17.1
-1.0
7.1
42.9
4.1
-
2.3
38.3
41.8
-
25.3
14.2
9.1
-
28.0
13.0
10.8
-
Whole
All levels
Children
Relative effect
(%)
Whole
Technical and
Technological
levels
Children
Other
institutions
Absolute
effect
Sample 1
Sample 2
Sample 3
Sample 4
Heads
S-system
Other
institutions
0.30
0.86
0.26
-0.23
0.60
0.35
0.25
-0.02
Children
S-system
Other
institutions
S-system
Other
institutions
0.46
1.60
0.62
-
0.57
0.49
0.58
-
0.55
-0.09
-0.22
-
0.51
0.28
0.09
-
Whole
Heads
Children
Relative effect
(%)
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
5.5
23.1
6.5
-5.5
10.5
9.0
6.0
-0.5
7.3
41.6
14.8
-
8.5
12.4
13.4
-
13.0
-2.4
-5.4
-
11.6
7.1
2.1
-
50
Table A.6: IPW average treatment effects of vocational training on formality likelihood by
institution, household position and aggregation samples (1-4), percentage
points and percentages over non-trained individuals.
Whole
Professional
qualification
Heads
Absolute
effect
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
0.09
0.09
0.06
-0.05
0.08
0.10
0.08
0.03
0.10
0.06
0.03
-
0.07
0.10
0.10
-
0.09
0.09
0.08
-
0.12
0.12
0.09
-
Whole
Heads
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
16.2
16.1
9.6
-8.1
13.4
17.4
13.9
4.4
17.6
11.3
4.9
-
12.4
17.7
17.6
-
16.8
17.6
13.4
-
20.8
22.0
15.7
-
Heads
Children
Absolute
effect
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
0.08
0.09
0.07
0.56
0.08
0.08
0.07
0.07
0.12
0.07
0.04
-
0.07
0.09
0.10
-
0.06
0.08
0.14
-
0.11
0.10
0.08
-
Whole
Heads
Children
Relative effect
(%)
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
14.4
16.5
11.4
76.8
12.5
13.7
11.6
11.2
19.5
13.3
6.1
-
11.0
15.1
15.7
-
10.5
15.2
23.0
-
18.9
15.9
-4.9
-
Whole
All levels
Children
Relative effect
(%)
Whole
Technical and
Technological
levels
Children
Heads
Children
Absolute
effect
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
0.09
0.09
0.06
-0.03
0.08
0.09
0.08
0.04
0.10
0.06
0.03
-
0.07
0.10
0.10
-
0.09
0.09
0.09
-
0.12
0.12
0.09
-
Whole
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Heads
S-system
Other
institutions
16.0
16.1
9.9
6.0
13.2
16.6
13.4
6.0
Children
S-system
Other
institutions
S-system
Other
institutions
17.8
11.6
5.1
-
12.0
17.2
17.2
-
15.8
17.2
15.0
-
20.3
20.6
15.1
-
51
Table A.7: IPW average treatment effects of vocational training on employment probability
by institution, household position and aggregation samples (1-4), percentage
points and percentages over non-trained individuals.
Professional
qualification
Absolute
effect
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect
(%)
Sample 1
Sample 2
Sample 3
Sample 4
Technical and
Technological
levels
Whole
Heads
Other
S-system
institutions
0.10
0.10
0.07
0.09
0.07
0.06
0.10
0.09
Whole
Other
S-system
institutions
17.3
17.3
12.3
14.6
11.2
9.4
20.4
17.1
Other
S-system
institutions
0.07
0.04
0.03
0.02
0.01
0.02
Heads
Other
S-system
institutions
10.2
5.3
3.2
2.5
1.0
2.5
-
Whole
Heads
Other
institutions
0.03
0.08
0.01
0.07
0.07
0.05
Children
Other
S-system
institutions
4.6
13.5
1.9
12.1
11.4
7.4
S-system
Children
Absolute
effect
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
0.10
0.07
0.04
0.02
0.12
0.09
0.08
0.10
0.09
0.03
-0.02
-
0.05
0.02
0.02
-
0.08
0.04
0.04
-
0.08
0.09
0.07
-
Whole
Heads
Children
Relative effect
(%)
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
16.9
12.2
7.0
3.1
20.0
15.5
12.1
19.0
12.9
3.0
-2.1
-
6.8
2.8
2.2
-
13.8
6.4
7.0
-
13.8
14.3
-3.7
-
Whole
All levels
Children
Heads
Children
Absolute
effect
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
0.10
0.07
0.07
0.09
0.10
0.09
0.06
0.09
0.07
0.03
0.01
-
0.04
0.02
0.02
-
0.03
0.02
0.07
-
0.08
0.07
0.05
-
Whole
Heads
Children
Relative effect
(%)
S-system
Other
institutions
S-system
Other
institutions
S-system
Other
institutions
Sample 1
Sample 2
Sample 3
Sample 4
17.3
12.3
10.5
17.6
18.1
14.7
10.0
17.6
10.5
3.2
0.6
-
5.7
2.5
2.4
-
6.1
2.7
10.7
-
13.6
12.6
8.3
-
52
Table A.8: IPW average treatment effects of S-system vocational training on the probability
of being employment in the same area of training, by household position and
aggregation samples (1-4), percentage points and percentages over trained
individuals in other institutions.
Whole
Absolute effect
Professional
qualification
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect (%)
Sample 1
Sample 2
Sample 3
Sample 4
Absolute effect
Technical and
Technological
levels
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect (%)
Sample 1
Sample 2
Sample 3
Sample 4
Absolute effect
All levels
Sample 1
Sample 2
Sample 3
Sample 4
Relative effect (%)
Sample 1
Sample 2
Sample 3
Sample 4
Heads
Children
S-system over other vocational training institutions
0.07
0.12
0.13
0.13
Whole
0.04
0.09
0.09
Heads
0.10
0.12
0.12
Children
13.4
34.8
36.5
40.0
6.8
18.3
18.0
-
28.0
35.6
36.6
-
Whole
Heads
Children
0.12
0.09
0.09
-0.05
Whole
0.07
-0.08
-0.06
Heads
0.10
0.07
0.06
Children
23.2
18.1
18.2
-11.3
12.4
-14.9
-11.9
-
20.7
14.6
12.8
-
Whole
Heads
Children
0.08
0.11
0.12
0.12
0.07
0.07
0.07
-
0.10
0.10
0.10
-
Whole
Heads
Children
15.9
30.8
32.0
33.2
11.3
15.2
15.4
-
25.1
28.7
29.1
-
53
Table A.9: Heckman estimation of selectivity corrected monthly labour earnings and the
impact of the S-system
Outcome: log labour
earnings
Variables
Outcome Equations
Selection Equations
Sample 2
Sample 3
Sample 4
Sample 1
Sample 2
Sample 3
Sample 4
-0.011
0.179***
0.181*
0.180***
0.547***
0.735***
0.519***
(0.040)
(0.046)
(0.097)
(0.019)
(0.040)
(0.051)
(0.082)
0.149***
0.424***
0.425***
0.316***
0.819***
1.059***
0.875***
Complete basic ed.
(0.050)
(0.049)
(0.104)
(0.024)
(0.047)
(0.058)
(0.091)
Incomplete
0.193***
0.501***
0.556***
0.421***
0.983***
1.203***
0.939***
secondary ed.
(0.054)
(0.051)
(0.105)
(0.029)
(0.051)
(0.061)
(0.094)
Complete secondary
0.253***
0.623***
0.636***
0.558***
1.190***
1.416***
1.262***
ed.
(0.056)
(0.050)
(0.104)
(0.023)
(0.047)
(0.056)
(0.087)
Incomplete
0.513***
0.941***
0.965***
0.663***
1.437***
1.601***
1.499***
secondary ed.
(0.071)
(0.064)
(0.119)
(0.050)
(0.090)
(0.097)
(0.145)
Complete secondary
0.807***
1.311***
1.287***
0.939***
1.919***
2.045***
1.856***
ed.
(0.071)
(0.059)
(0.114)
(0.033)
(0.064)
(0.070)
(0.102)
Average household
0.036***
0.027***
0.033***
-0.016***
-0.031***
-0.034***
0.002
ed.
(0.003)
(0.003)
(0.004)
(0.002)
(0.004)
(0.005)
(0.007)
0.088***
0.106***
0.061**
0.101***
0.129***
0.127**
0.183***
S-system
(0.023)
(0.021)
(0.031)
(0.030)
(0.049)
(0.053)
(0.058)
-0.081*
-0.035
0.199***
0.135*
0.104
S-system x FEMALE
(0.042)
(0.036)
(0.044)
(0.073)
(0.078)
0.046***
0.065***
0.061***
0.047***
0.075***
0.079***
0.075***
Potential experience
(0.004)
(0.004)
(0.007)
(0.001)
(0.006)
(0.007)
(0.010)
Potential experience
-0.001***
-0.002***
-0.002***
-0.001***
-0.001***
-0.002***
-0.002***
(sq.)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
(0.001)
-0.086***
-0.097***
-0.083***
Union Membership
(0.010)
(0.009)
(0.016)
-0.095***
-0.274***
-0.902***
-0.809***
-0.768***
Women
(0.026)
(0.017)
(0.013)
(0.022)
(0.021)
Children in
-0.018***
-0.022***
-0.030***
0.012***
-0.025***
-0.027***
-0.045***
household (n)
(0.005)
(0.005)
(0.009)
(0.004)
(0.007)
(0.009)
(0.013)
Bi-parental
-0.018
0.016
0.025
-0.168***
household
(0.013)
(0.018)
(0.023)
(0.032)
0.051***
0.049***
0.022
-0.071***
-0.027
-0.001
0.040
Ethnicity 1 (white)
(0.018)
(0.017)
(0.031)
(0.022)
(0.032)
(0.035)
(0.049)
0.007
-0.018
-0.048
-0.051**
-0.056*
-0.067**
-0.068
Ethnicity 2 (mulatto)
(0.017)
(0.017)
(0.032)
(0.021)
(0.030)
(0.033)
(0.047)
-0.022
-0.166***
-0.339***
Rural
(0.023)
(0.019)
(0.027)
-0.478***
-0.161
0.023
Inverse Mill’s ratio
(0.044)
(0.025)
(0.038)
6.576***
6.025***
5.942***
0.237***
-0.409***
-0.608***
-1.152***
Constant
(0.108)
(0.100)
(0.194)
(0.033)
(0.061)
(0.071)
(0.106)
Occupation controls
Yes
No
Sectoral controls
Yes
No
State controls
Yes
Yes
Observations
98077
33027
25788
12235
98077
33027
25788
12235
Population size
47670593
15548981
11988426
5685453
47670593
15548981
11988426
5685453
Censored population
20,703,397
6,430,786
4,492,326
2,859,725
20,703,397
6,430,786
4,492,326
2,859,725
F
467.92
98.098
122.04
59.50
467.92
98.098
122.04
59.50
Prob > F
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Significance levels: *** p<0.01, ** p<0.05, * p<0.1.
Note: Based on sample 1 (whole population aged over 10 years). Excluded categories: no education, agriculture, managing directors and
CEOs, other ethnic backgrounds.
Source: Own calculations based on PNAD 2007
Incomplete basic ed.
Sample 1
0.205***
(0.016)
0.421***
(0.018)
0.503***
(0.021)
0.626***
(0.019)
0.925***
(0.031)
1.311***
(0.027)
0.034***
(0.002)
0.093***
(0.014)
-0.013
(0.025)
0.037***
(0.001)
-0.001***
(0.000)
-0.077***
(0.006)
-0.455***
(0.010)
-0.012***
(0.003)
0.096***
(0.011)
-0.006
(0.011)
-0.126***
(0.016)
0.009
(0.011)
6.126***
(0.050)
54
Quantile Regression Methodology
Consider the following distribution function from a random variable
)
For any
(1)
the quantile of
{
is defined as:
}
(2)
The quantile function gives a complete characterization of . Q(0.5) is the median, the first
quartile is Q(0.25), the first decile is Q(0.1) and so on. We can now write it in a different way.
Let
a
vector of explanatory variables, then:
|
|
(3)
This equation is equivalent to:
(4)
Where the only constraint is that
|
and the distribution of this error term left
unspecified. From this we now write the linear conditional quantile function as:
|
Where
(5)
can be estimated solving:
∑
(6)
55
(
Where
) is the piecewise linear “check function” and
is the
indicator function. The quantile regression estimator achieves a more complete description
of the conditional distribution of
quantile of
given
. The partial derivate of the conditional
with respect to one of the regressors, j-th, could be interpreted as the marginal
change in the
quantile due to a marginal change in the j-th regressor. Note, that this
marginal effect is related with and not with some particular individual, which changes their
quantile simultaneously when some value of their explanatory variable changes.
There is a major advantage using Quantile regression. It can detect and deal with the
plausible heteroskedasticity, as it is present in the Brazilian data allowing different slopes for
each conditional quantile in the wage distribution. Using the method suggested by Koenker
and Bassett (1978, 1982) to obtain the standard errors, Rogers (1992) reports that these
standard errors are suitable in the homoskedastic case but that they look to be understated
in the presence of heteroskedastic errors.
The median regression is also more robust than an OLS regression in presence of
outliers and a difference of Heckman ML estimators, it does not rely on the normality of the
residuals because the estimation is only influenced by the local behaviour of the conditional
distribution near the specified quantile.
56

The Impact of SENAI`s Vocational Training

Transcription

Similar documents

Vocational Scholarship Program

MENA Workshop İstanbul January 2013 «Active Employment

Going for Goals with the MKI-DS - (EPP) Egypt

Centre for environmental technologies

Georg - Kerschensteiner Schule

Local Labour and Materials - Frontiers Foundation Manitoba

Contact persons - ADAM - Leonardo da Vinci Projects and Products

Labour Market Information

kanpur urban pilot project - World Vision International

S011 Construction sector leaflet